If you're going to put a video demo on your main webpage, can it have play/pause and a control bar? So I can actually skip to a part I want to look at. Here's the actual video: https://tui.studio/screenshots/video.mp4.
Also, how does this handle terminal resizing? Are there options to anchor elements to the left/right etc, or will narrowing the terminal window just make everything fall off the side, or worse, all the text wraps?
> If this attribute is present, the browser will offer controls to allow the user to control video playback, including volume, seeking, and pause/resume playback.
Edit: I misunderstood, you are asking
> how they'd managed to hide the video context menu
Its entirely possible I did something to it accidentally that made the context menu not work properly. I had the dev tools open to pull the actual video address when I right clicked, so I might have messed something up. Or maybe the devs are secretly looking at the comments and fixed it between me and you trying :P
This is nonsensical, there is nothing textual about the UIs being shown here. It doesn't stop being a GUI if you have a 1:1 representation of the concept within character cells.
The UX actually matters, and TUIs are generally built for effectiveness and power (lazygit being an excellent example). But once you start adding mouse clickable tabs, buttons, checkboxes etc. you left the UX for TUIs behind and applied the UX expected for GUIs, it has become a GUI larping as a TUI.
> But once you start adding mouse clickable tabs, buttons, checkboxes etc. you left the UX for TUIs behind and applied the UX expected for GUIs, it has become a GUI larping as a TUI.
Hard disagree. Borland TurboVision [0] was one of the greatest TUI toolkits of the DOS era, had all of these:
> Turbo Vision applications replicate the look and feel of these IDEs, including edit controls, list boxes, check boxes, radio buttons and menus, all of which have built-in mouse support.
Vasellating. TurboVision was awesome, but it was pushing the boundary of TUI, which in my mind was great for moving hard copy to computer entered use case. To wit, hard copy on your right side, you transfer data to app without looking at screen, but just looking at hard copy, remembering when/where to hit return key, maybe tab for prior field, stuff like that.
But hey, if the screen is drawn 24 x 80 with extended ascii, it's TUI. And man, loved the "absolute" keyword in turbo pascal. Instant screen writes when writing to a 2 dimensional array.
It's a TUI if it uses text to build those elements.
You can be effective and powerful in any kind of interface, Just like you can be ineffective and weak in any kind of interface. People like TUIs because they're cool, and work over SSH.
Yes.
A TUI runs in a text session.
A GUI runs in a graphics session.
A terminal emulator emulates a text session in a graphics session - and allows you to run TUI/CLI tools.
This is apparently controversial?
> TUI runs in a text session. A GUI runs in a graphics session
What do you mean by this? I have never heard these terms before. I can launch and interact with a GUI from a text application, or a text application from a GUI.
It makes far more sense in the context of effectively a remote session or dumb terminal serial connection. Your "terminal" application is only emulating a text mode environment inside a gui, typically... you can ctrl+alt+F2 - F5 in a lot of Liknux's to switch to a terminal session if you want, but that's not what people tend to actually use.
Beyond this, without remote X properly configured, again, most don't and probably shouldn't.. you aren't running remote gui applications over an SSH session. Richer TUIs were pretty common in ye old days of DOS and other OSes before rich GUIs become the norm. DOSShell, Edit.com, etc. The IDEs of days past and Word Perfect even. These all interacted with Mice and were considered the norm. The features that allow this over a remote terminal today are pretty great IMO, the harder part is properly handling window sizes/resizes, etc.
With graphical extensions, there are even nice app explorers with image previews via TUI. It pushes the boundaries. For that matter, I often wonder what could have come with RIPscrip/RIPTerm if the leap to web didn't happen the way it did...
I think the single hardest part of TUI is dealing with wide characters and secondary fonts for color emojii that don't quite render in 2 spaces completely in a lot of termianls... it makes the line drawing harder too.
In the Linux world, your GUI is built on a stack of separate programs. At the bottom is the "display server", such as X11 or Wayland. One of these needs to run to have GUI windows (or a full-screen background) at all; otherwise the screen is just covered by 80x24 terminal, and even if there were mouse support there would be nowhere to click or drag to change that. Without the display server, you are in a "text session". It's relatively rare to do this locally nowadays, but Linux still adheres to the highly modular UNIX philosophy.
Your terminal windows (whether that's "Terminal" or "cmd.exe" or anything else) are still fundamentally graphical programs that emulate such a text session.
Is it character mapped, designed to run in a tty terminal? TUI.
Is it pixel or vector mapped, designed to run in a graphics terminal? GUI.
Of course strictly speaking TUI is a subset of possible graphical user interfaces, but the term GUI was coined to denote interfaces other than the already-ubiquitous text terminal interfaces.
TUIs have since absorbed GUI interface elements like buttons, checkboxes, and even pointer input, which I think is causing the terminology complaint here. Classical TUIs like Norton Commander are more about keyboard input and navigation. But being text-mapped is the identifying feature of a TUI, I think most people accept.
Sometimes your system doesn’t have a graphical session, like a raspberry pi with no x server running, or a cloud compute instance I’m ssh’ed into, or a docker image running on my laptop. Sometimes your system doesn’t have a (particularly usable) text system, like a work computer that disables the terminal or a family member’s MacBook who doesn’t have the time or space to install XCode terminal utilities to be able to use things like brew install.
My point is that it’s not a given that having one means you have the other.
Perhaps he's thinking about "console" / "display server" but the lines blur fast (e.g. you can run GUI in linux console with framebuffer with some limitations)
Of course you can use the primitives of TUI, especially with mouse support, to reproduce a large amount (if not all) of the standard GUI interaction paradigms.
But it's bizarre, and missing the point from a UX perspective.
As an extreme example, we can imagine a program that displays the borders of a 40x15 "window" in the middle of a console, with box-drawing characters, putting a "close box" in an upper corner, with text like "File Edit Help" in the top left. We can imagine it responding to a click on the "File" text by popping out a "menu"; we can imagine a drag starting from the "title bar" causing the window position to be update (and the entire terminal window redrawn).
A lot of those kinds of functions, ironically enough, might make sense for a TUI editor implemented as a TUI (except the "windows" might just be understood as panels where the ultimate program displays parts of its output). But as an emulation of GUI windows, it'd be a strange, impractical novelty.
What's with the purism? It's just a term used to differentiate one way of making a UI from another. Who cares about what is practical when you're just trying to give a thing a name.
Even in your example, it's pretty clear cut. If the window is built with text and served in a terminal emulator, it's a TUI. If you build it with a graphical framework that now needs X11 or whatever, it's a GUI.
I assume his point is that making stuff that assumes a mouse makes for a bad text-based UI. Absolutely fine if everything is controllable via the keyboard, e.g. if the tabs were labelled F1-Fn and they function keys switched them, or they had an underlined letter and Ctrl+letter switched focus to it, or whatever.
But if this thing requires you to just tab a lot through lots of pointless and rarely used fields to get to a "button" so you can activate it, because it's really all designed to be used with a mouse, then it's a bad text-based UI.
There are some incredibly good text-based UIs around, some going back to mainframe stuff from the 70s. Most of them are optimised for speed of control via keyboard rather than for looking pretty. Almost none of them would be quicker to use with a mouse.
absolutely, but it's still a TUI. Just like if you made a GUI that didn't have any mouse support and needed all keyboard shortcuts, it wouldn't stop being a GUI.
TBH I've always hated the backronym TUI anyway, so if TUI includes things that require mouse input, then maybe we can go back to using "text-based UI" for things that only require a keyboard like we used to 30 years ago.
Would you say old DOS applications like Borland's Turbo series of compilers were not TUIs? They ran in the console but had menus, mouse support, dialog boxes, etc...
How about those text games that used ASCII art and you typed in commands like "look" and "go north"?
I would say using text mode is the primary requirement for a TUI. The other requirement being some kind of human-machine connection, IE a User Interface.
> there is nothing textual about the UIs being shown here.
Well, except:
> a 1:1 representation of the concept within character cells.
TUI is build from text, and living within its constraints and what it's engine (usually the terminal) allows. GUI is build from graphics, and has basically a pixel perfect control of its own. This is a very notable difference, especially at the time when these terms were coined.
> TUIs are generally built for effectiveness and power
No, this is a result of different architectures and their constraints.
> But once you start adding mouse clickable tabs, buttons, checkboxes etc. you
TUI and mouse are predating the GUI (more or less). We had them already 40-50 years ago at the dawn of interfaces. We are now just moving back to them for practical reasons.
The UIs are text only, so they are textual. Modern TUIs may support mouse events. That this tool can export to several TUI frameworks is evidence that these UIs are indeed TUIs, even if not the most traditional.
Exactly, don't know why people are acting like actually makes TUIs, it's just a rough mockup of a TUI for now, with a convoluted figma-like UI.
I guess the headline and website was enough to get all these upvotes. Quite disappointing as someone in the early stages of making a TUI tutorial myself.
I've been juggling some BBS related projects myself that involve some TUI work over raw and web sockets that I've been working on... It's definitely a fascinating space and there's been a lot of relatively recent activity in the space.
> and there's been a lot of relatively recent activity in the space.
100 percent agree. I personally love what the openTUI folks have been up to. As weird as this might be to say, we're still in the early, early stage of TUI adoption.
Zellij among is a great example, I can do everything with my keyboard, but every now and them I'm already with the mouse and just click a tab or pane, no functionality lost, just added, why the need to make a cutoff philosophical/semantic hard argument?
Good insight, but if you discount the visual elements (tabs, buttons, etc), you're limiting TUI to CLI, and I think that's unwarranted. The value proposition of both TUI and GUI is two-fold: you see the available action options, and you see the effect of your actions. So, yes, TUI and GUI _are_ closely related: who cares whether we're displaying pixels or character blocks.
Unfortunately, they are often artificially differentiated by the style of the UX interaction: TUIs promote the keyboard actions, and GUIs prefer mouse without corresponding keyboard shortcuts. Unfortunately for GUIs, their designers are often so enamored with WIMP that they omit the keyboard shortcuts or make them awkward. I hate it when, even if the ACTION button is available by keyboard traversal at all, it requires some unknown number of widget traversals instead of being one tab away.
Since the keyboard is almost always used for the textual data, it makes sense to me to always enable it for command execution. Well designed GUIs and TUIs provide both WIMP and keyboard UX, which sadly is not the norm today, so here's my vote to make them larp for each other more.
Would you make the same argument for classic UIs created with things like Borland's Turbo Vision framework? It's generally known as a TUI framework (including by Wikipedia).
That's fair... I feel that way about GUIs too in general though. Everything should be keyboard navigable and reasonable control flows. Tab and arrows, etc. Should be able to control focus and selection (enter).
I admit I don't always pay the most attention to it, as the UI components I tend to use do a good enough job of this. But I'm usually pretty consistent with it.
You can tunnel a port over SSH and get a web UI locally, though it's not commonly done. I feel like more people would actually do this if tunneling a port was just ever so slightly easier (like, you're already SSH'd into a box, then you run a command, then you somehow automatically get a tunnel for that command's UI port plus a local browser window open to the page)
While in an SSH session, press enter, then type tilde and capital C (enter ~C) and you can add command line options to the current session. To add a port forward from your local 8080 to the remote port 80 without closing the connection, do:
SSH expects the escape sequence (tilde) to be the first character on a new line; since backspace is sent as a character, you can't just backspace over something you've started typing and then press tilde to have it recognized.
Technically, you don't have to press enter if you've not typed anything (try it in a new SSH session - as soon as you are logged in, type ~? to get the SSH help output), but since the comment was about doing this during an active session without ending it, I figured noting that pressing enter first to be sure you're on a new line wouldn't hurt
share some good (easy on remembering keyboard & mouse) tmux configs. I usually struggle with copy pasting many scrollback lines from/to tmux. would love for my claude to be natively tmux aware.
I do this a lot but I'd still prefer TUI where possible. With too much visual content it isn't of course, but for many cases a TUI is much more responsive and much lower resource.
I largely agree with you, but there are limits to what a tui can do well. If analyzing a flame graph or performance trace, web UI is a better fit. However, most things are not that.
Even easier is just using an X server, if you have it set up properly you just need to run the remote app and the window pops up on your machine.
(I think terminal-based GUIs are neat just for fluidity of use- you can pop one open during a terminal session and close it without switching to mouse or shifting your attention away from the terminal. They can also be a nice addon to a primarily CLI utility without introducing big dependencies)
Yeah I love that about X. I remember in the 90s when I first figured that out. I was logged in from a university workstation into my home computer with SSH and I launched my mail client or something and I thought doh, stupid that will only popup locally.
Then colour my suprise when it popped up on my screen right there. Slow as molasses but still. Wow. Magic.
It's a shame Wayland dropped this. Yes I know there's waypipe but it's not the same.
It... really isn't. Like you said, remote X was barely usable even over an entirely local network. Most applications these days are also not designed for it, using loads of bitmap graphics instead of efficient, low-level primitives. So you end up being just one tiny step away from simply streaming a video of your windows. We have better tools for doing things remotely these days, there's a reason approximately no one has used remote X after the mid-90s. It's a neat party trick, but I don't blame the Wayland authors for not wanting to support it.
> one tiny step away from simply streaming a video of your windows
In the 80s/90s this wasn't feasible due to network latency and bandwidth, but it's pretty common now to do exactly this, with VNC and other remote desktop protocols.
The big obvious one is web-based tooling. Your information & settings are stored on a server and you use a web browser to view it via whatever device you're on. For more locally based workflows, we have networked filesystem protocols, automatic syncing between systems, that kind of thing. It's not a 1-1 equivalent of running a remote program and viewing it locally obviously, but it gets the same job done, in a much more useful & flexible manner than X forwarding did.
For example, the remote mail client usecase I was replying to is simply done with a webmail client today.
I don't really feel like web interfaces or syncing are really a substitute tbh, and I'm not sure how they're more flexible. ssh -> run -> gui opens, and the program itself doesn't need to be designed differently to work
> and I'm not sure how they're more flexible. ssh -> run -> gui opens
But this doesn't work on your phone, or on a Windows or macOS device, right? That's what I meant by flexible, X forwarding fits a pretty narrow set of usecases, while on the other hand keeping programs on the clients and data centrally located on a server allows for a whole lot more options for how to interface with that data.
(To be clear, nothing wrong with X forwarding! It's a cool tech and I'm glad you have a use for it! I'm just arguing that it's fine for Wayland to not try to support that kind of thing, because we've got other ways of working remotely now.)
X servers are available for phones, Windows, and macOS. X interfaces not designed for phones can be difficult to use on phones. But web interfaces not designed for phones can be difficult to use on phones.
There is not a web tool for every use. And web tools are not better for every use.
The main advantage of x forwarding for me was when I'd randomly need it and had nothing set up ahead of time. Hopefully it starts getting installed in distros by default eventually.
People don't build TUIs because they want to run apps in the terminal, they build them because the terminal happens to be the most portable app platform available.
I've been working with notcurses recently and it is a full TUI that handles mouse events just fine. Runs over slow SSH connections and everything. The nice part is that you can fully operate applications built on top of it with the keyboard if you so choose, the mouse is just a shortcut.
Sadly the project is not really in a usable state at the moment. The documentation is incomplete riddled with errors, the code has some pretty glaring bugs, and it's close to abandoned. It's a shame because you can do some really amazing stuff with it.
You might not like this type of interface, but it is hardly "nonsensical". In the 1990s this sort of text-based GUI was common in DOS programs, such as Borland's "Turbo" languages and the original pre-Windows FoxPro.
lazygit supports vim style keybindings and mouse click and scroll. I mostly use the key shortcuts but sometimes the mouse is useful. But i agree that a well thought out state machine that can be navigated through via keyboard is a dream to work with. Lazygit is superb. But this is not a distinction between TUI and GUI.
> But once you start adding mouse clickable tabs, buttons, checkboxes etc. you left the UX for TUIs behind and applied the UX expected for GUIs, it has become a GUI larping as a TUI.
TIL that VIM is not cease being TUI the moment I type :set mouse=a.
Hot taking, LARPing and teenage angst (caused by generational gap with those has been using TUI since 1980s) is on your side.
My ancient boxed copy of Visual Basic for DOS 1.0 that supported mouse clicks on TUI buttons would have found your viewpoint quite offensive if it had any AI in it ;-) Oh boy, good old days.
Man, I've had so much frustrating just trying to copy & paste from inside a terminal running e.g. opencode or crush.
I think TUIs are neat, I guess. But I think these things have abused the concept extensively. They don't actually interact well with the rest of a Unix environment.
Drawing a “nonsense” line between TUIs and GUIs is pretty arbitrary, it’s all pixels on a screen at the end of the day. People like the TUI vibe, and that’s a good enough reason to make and use them.
I love TUIs but one main reason for that is that they're keyboard centric. If I have to use the mouse it kills it for me, if both work then it's fine. I hope that modern TUI makers keep this in mind. What's great about the keyboard centric is that with a few keystrokes/shortcuts it's very easy to do repeatable work and takes less energy than hunting boxes to click on with the mouse.
TUIs aren't more inherently keyboard driven than well constructed GUIs. You can easily make a keyboard driven GUI that has all the shortcuts you'd add to a TUI. (Just don't let the "UX design experts" near it.)
I really don't want my TUI's to look like GUI's rendered in low res. The appeal to me of a TUI is that it is built specifically to be a TUI, and that means eschewing complexity and detail, and favouring compact text.
That's literally what TUI's looked like starting from the late 1980s and throughout the 1990s... You have a pointing device, might as well make use of it to enhance discoverability.
This seems really reductive. Some UI paradigms work better at 80x24 vs at 640x480 (never mind whatever resolutions we have access to today). Or rather, the 80x24 text grid is using more pixels than that, but everything is aligned to that lower resolution, and that fundamentally changes what makes sense to do. Floating windows that can be dragged around to arbitrary positions? Terrible for low-res; classic for higher res. Dividing lines that split the screen into panels, and can be moved around a row or column at a time with a keyboard shortcut? Pretty much the opposite (enthusiasts of "tiling WMs" might disagree).
This is why I don't like TUIs at all, they're really bad at displaying complex information, handling complex interactions, and discovering how to compose those together.
> Design once, generate production-ready code for your framework of choice. Switch targets without touching your design. Alpha notice: Code export is not functional yet. We're actively working on it — check back soon.
In other words, it isn't at all usable right now. You can't produce a TUI with it, not even a limited one.
Ignore the haters, sure. But don't ignore the well argumented criticism that you're getting from an overwhelming majority of your peers, as it's happening right now.
>can't wait to see where it goes!
Fall into this toxic positivity nonsense at your own peril.
Gatekeeper blocks the app immediately. You'll see either "TUIStudio cannot be opened because it is from an unidentified developer" or "TUIStudio is damaged and can't be opened" on newer macOS after quarantine flags the binary.
To get past it: right-click the .app → Open → Open anyway — or go to System Settings → Privacy & Security → "Open Anyway".
A trusting, highly positive person could really be taken advantage of here.
Better get ready for almost all software to use AI assistance in its creation.
You can build great things using AI agents, and you can build trash.
Your ideological opposition to that is not shared by as wide a percentage of developers as you may think based on some highly self selected online corners.
Way to read something into what I have never even wrote and try to spin it as something ideological. This is not "AI-assisted", it's completely vibe-coded by AI and the software doesn't even make sense since you can't export anything. It's just low-quality trash dumped on Hacker News and I'd argue this is not the place for it.
It's not anything though. It's a website and electron app that promises functionality that completely isn't there. It's useless, but instead of being art, it promises functionality, so it's functionally trash.
I too remember running `rails new MyGreatApp` and having hoop dreams of being the next billionaire entrepreneur, but a boilerplate app is a boilerplate app.
Your ideological opposition to that is not shared by as wide a percentage of developers as you may think based on some highly self selected online corners.
Opinions do not win by "high score". Asserting the validity of opinions based on how widely they are held is dumb.
If 30% of devs love openclaw, I'm not re-examining my informed opinion, I am forming a new one about that 30% cohort.
Apparently we now write desktop applications intended for designing the UI of other desktop applications, in TypeScript that runs in a Docker container, using a bunch of web frameworks (including for CSS) and self-hosting an nginx server?
I would have expected a TUI editor to be itself a TUI.
Funny how you can tell a project is vibe-coded just from a first glance at its website. All these websites seem to somehow have the same visual style. Anyone noticed this?
Yep, my fans started revving as soon as I loaded it. Animations are out of control on the normal web as it is, but genAI sites take it to another level.
Yeah it was. I've been hoping to get the time to write a userscript to identify and stop animations like that, but I haven't been able to understand the whole requestAnimationFrame thing enough yet.
TUIs are very in right now. Nostalgia/vintage-computin-aesthetics & guru-gatekeeping around command line savviness are front and centre in the HN-and-adjacent mindshare.
I agree. The animation on the site lost me when it placed a button. IMO, buttons are not part of TUIs. Those are just low-resolution GUIs, IMO, and that’s sort of the worst of all worlds. The first good TUIs were things like top and elm.
ENSHT comes for everyone. This is sexual selection over natural selection. Claude Code also gets this wrong, they got way to fancy and ruined what a good tui is by being an uncanny combo between a scrolling log and a completely rewritten canvas.
In this age, rich TUI's feels wrong to me. Tools that expose a minimal web server with a lightweight UI is much more welcome than a complex TUI. But for most interactive terminal apps, it feels more natural when there is a single input at a time, like a wizard interface.
This looks really cool. However, the current AI models are pretty good at designing UIs from prompts and even turning screenshots of mocks into full UIs. I'm not sure this visual design approach would save time vs simply prompting an AI agent.
That being said, I could see a niche market for a designer persona who is used to building in tools like figma.
also, this assumes humans are still the primary CLI consumers. With agents increasingly being the first-class users of command-line tools, building visual design tooling for terminal UIs feels like optimizing for a shrinking audience.
On one hand this is a neat idea. I've thought about how nice it would be to have a visual layout tool for text-based designs. The current offerings are slim. Of course, you could easily argue that if you need a visual tool for it, you've gone too far; even the most sophisticated TUIs are still extremely simple.
On the other hand, for this work as they describe, it needs to be a complete UI framework across a bunch of languages and built on top of a bunch of existing frameworks. That seems... ambitious. Building one UI framework for one language is plenty hard enough.
The TUI hype seems like nostalgia for COBOL mainframe apps that most people have never even used. A sort of secondhand cyberpunk role play with zero focus on actual UX.
Also if TUIs are so great, why isn't this a TUI app?
Have you ever watched someone USE those COBOL TUIs? Everyone from airline ticket agents, to local governments, to folks at Home Depot while looking up inventory. They could fly through menus and accomplish things. I remember when Best Buy switched to a Windows-based experience. It was terrible. Simply adding a mouse+windowing experience slowed everything way down. I saw it first hand at Target too. They went from an OS/2-based TUI to Windows NT. I know there'll always be those folks that think we're all just trying to play "leet Haxorz", but there's just something about those systems that people deeply connected with.
I personally think the opencode and kilo CLI have great UX, certainly better (and easier to use) than the web versions of both.
A lot of the recent TUI apps are really not old-school in any way. Not all apps need the feature-set of a browser engine. And compared to native mac/linux desktop apps, TUIs get cross-platform support by default.
> Also if TUIs are so great, why isn't this a TUI app?
Half surprised there's no raratui export with the other options. That said, probably lends itself more to Ink and @opentui/react. Also slightly disappointed at the lack of a direct Linux build for AppImage and/or Flathub. Also not using Github's releases which is a little curious.
This is a vibe coded app and isn't what I'd want but still, it's interesting to consider what a good implementation of "Figma for TUIs" could be, especially if it avoids the trap of simply treating the console as a crude raster instead of taking advantage of text and keyboards. IMO we don't need WIMP GUI shoved into terminal emulators.
Depending on how the TUI is made, it can be very visual, but lacks structure for a screen reader (unless you stay in the very simple "input field: value" kind of prompt, but even then auto completion is tricky).
Web browsers offer the DOM to tools such as screen readers (OSs offer their own accessibility sdks). Someday perhaps the TUI application could talk to the terminal emulator that would itself talk to the accessibility sdk of the OS and that info would somehow then be accessible.
There was a beginning of discussion at bubble tea[0] about this for example.
TUIs seem like they should be well-suited to screen readers, but yeah accessibility is often unfortunately far down the list on average project priorities.
The TUI-vs-GUI debate in the comments is probably not the most useful frame. The practical question for something like this is whether the exported code models terminal layout constraints correctly. Textual and Ratatui both use constraint-based layout, not absolute positioning, and a designer that works in pixel coordinates will produce UIs that look right at one terminal size and fall apart at others. Has anyone tested an export against real usage?
TUIs built today should be usable by AI agents. I'm not sure exactly what it looks like but I'm imagining that every UI view has an associated CLI command that can yield precisely that view. Maybe like formally structured breadcrumbs, or maybe like Emacs "keyboard macros".
I've found agents effective using GUI apps with nothing but the ability to take screenshots and send mouse and keyboard commands. I imagine they'd work even better with a TUI, even if it's not designed with agents in mind at all.
This is so cool I immediately wanted to convert my apps. But then when I thought about it, well it's trying to recreate CSS but in a majorly worse way.
Browsers are ubiquitous and I can just tell ai to build a web page.
I can't really see a use case other than novelty.
If you need a UI over SSH or inside tmux, skipping browsers and CSS isn't just a novelty, it's essential since HTML can't touch that territory. TS-based layout in terminals can be ugly but it also dodges a pile of accessibility, latency and bloat issues you get by default with anything running in Chrome.
The biggest pain point with TUIs has always been the design iteration loop. You're basically writing code blind, running it, tweaking numbers, running again. It's like writing CSS without a browser preview.
Something like this could genuinely help for the layout/positioning phase, even if you still hand-write the interaction logic. The debate about whether these are "real TUIs" kind of misses the point imo. Textual and Ratatui already blur that line with mouse support and rich widgets. The ship sailed on pure keyboard-only text interfaces a while ago.
What I'd actually want from a tool like this is to export to multiple TUI frameworks. Right now you're locked into one ecosystem and the code export isn't even working yet, which makes the whole thing feel premature.
I built something like this in 1993, it was used to design layouts for DOS apps and headers for printed listings. Imitating the BorlandPascal and Novel TUIs of the day
I'm not sure the utility of this kind of stuff anymore. It's relatively easy to sketch a layout on a napkin + prompt and then prompt claude code to use python textual as as TUI layer. I've had pretty good success with Textual+Claude so have a few colleagues. You could probably use Figma + claude etc as well.
The hardest part of TUI development has always been the layout model, not the widgets. Most people start with absolute positioning and then discover their terminal is 80x24 on the CI server. A visual designer is useful if it outputs layout constraints rather than fixed coordinates. Otherwise you end up with UIs that only work at one terminal size.
> No install fuss — download and start designing immediately.
also
> Gatekeeper blocks the app immediately. You'll see either "TUIStudio cannot be opened because it is from an unidentified developer" or "TUIStudio is damaged and can't be opened" on newer macOS after quarantine flags the binary.
To get past it: right-click the .app → Open → Open anyway — or go to System Settings → Privacy & Security → "Open Anyway".
Claude Code built a TUI for me last night, in this case to step through nanosecond timestamped ITCH market data messages and rebuild an order book visual in the terminal. This type of stuff would have taken a day - but done in 5 minutes now.
I would be REALLY REALLY impressed if it manages to do this without bugs. Just using pythons textual can be very complex, belive it or not. Maaging not only to that but other frameworks too sounds insanely complex. I have a strong feeling this is vibecoded from the commit history?`
Ah yes, it says clearly that on the github page. Still, if its works, I am then impressed by the LLM.
Edit: It does, in fact, NOT work for code export. Level of impressiveness massively dropped.
Yeah, the website has many bugs too. Literally can't click on 50% of the "clickable" stuff. Not impressed by vibe coded nonsense. The comments here are weird, people are discussing the "idea" rather than the broken implementation.
Probably a bad omen of things to come for the internet.
Am I the only one who thinks the recent TUI explosion is absolutely not necessary?
I mean yes, code editor are great for this but a lot of the TUIs I see are so slow it begs the question why they exist to begin. CLIs are supposed to be remixable and scriptable.
I think a better architecture would be to generally keep CLIs work like CLIs and have separate processes that add terminal rendering functionalities for those that need / want it but in general it is an anti-pattern to start from this as default.
Depends highly on the specific application. Take a simple example of looking at process usage. You can use ps from the command line to get all sorts of info about a process. But there’s no substitute for top to show you an updating list of top cpu consumers, which ps just can’t do.
That's roughly aligned with my thinking. Make it a CLI. And if there's a lot of configuration that you can pass to it, have an option for rendering those options as a TUI.
Of course it's not necessary, it's a fashion. Choosing to make a TUI instead of a GUI is a fashion statement, it signals aesthetic alignment with nerdy shit and says the program isn't meant for common proles. There's pretty much nothing a TUI can do that a GUI can't do, while the opposite is very much not the case.
I wish HN had flairs (tags) like Reddit and mandated a few for AI-related work (AI-Assisted, AI-meta, AI-vibecoding) or something so these could be filtered out
I'm not sure that's a fair criticism. Many things require or benefit from something even more complex to make them (car -> factory, code -> IDE, text -> editor, food -> kitchen). I think the real debate here is that which is found in the other comments: do we want TUIs to look like GUIs?
The only reason I can think of is what I said before [0] given that the web was destroyed by the same web developers, then so was the desktop (with Electron) and now of all places terminals are now getting destroyed with infinite slop like this.
This nonsense will continue and accelerate until it reaches hardware.
If you're going to put a video demo on your main webpage, can it have play/pause and a control bar? So I can actually skip to a part I want to look at. Here's the actual video: https://tui.studio/screenshots/video.mp4.
Also, how does this handle terminal resizing? Are there options to anchor elements to the left/right etc, or will narrowing the terminal window just make everything fall off the side, or worse, all the text wraps?
You can right click on it and choose "Show controls", at least in Firefox.
Oh, that's odd, it didn't show up on chrome when I first tried it, but it does now. I was wondering how they'd managed to hide the video context menu
It's probably just <video> element without "controls" attribute.
https://developer.mozilla.org/en-US/docs/Web/HTML/Reference/...
> controls
> If this attribute is present, the browser will offer controls to allow the user to control video playback, including volume, seeking, and pause/resume playback.
Edit: I misunderstood, you are asking
> how they'd managed to hide the video context menu
Not sure, but it works in FF for me
Its entirely possible I did something to it accidentally that made the context menu not work properly. I had the dev tools open to pull the actual video address when I right clicked, so I might have messed something up. Or maybe the devs are secretly looking at the comments and fixed it between me and you trying :P
It won't let me reply to parent's child comment, but i wanted to say:
That is what HN is for!
This is nonsensical, there is nothing textual about the UIs being shown here. It doesn't stop being a GUI if you have a 1:1 representation of the concept within character cells.
The UX actually matters, and TUIs are generally built for effectiveness and power (lazygit being an excellent example). But once you start adding mouse clickable tabs, buttons, checkboxes etc. you left the UX for TUIs behind and applied the UX expected for GUIs, it has become a GUI larping as a TUI.
> But once you start adding mouse clickable tabs, buttons, checkboxes etc. you left the UX for TUIs behind and applied the UX expected for GUIs, it has become a GUI larping as a TUI.
Hard disagree. Borland TurboVision [0] was one of the greatest TUI toolkits of the DOS era, had all of these:
> Turbo Vision applications replicate the look and feel of these IDEs, including edit controls, list boxes, check boxes, radio buttons and menus, all of which have built-in mouse support.
Well, I can’t remember if it had tabs.
[0] https://en.wikipedia.org/wiki/Turbo_Vision
Vasellating. TurboVision was awesome, but it was pushing the boundary of TUI, which in my mind was great for moving hard copy to computer entered use case. To wit, hard copy on your right side, you transfer data to app without looking at screen, but just looking at hard copy, remembering when/where to hit return key, maybe tab for prior field, stuff like that.
But hey, if the screen is drawn 24 x 80 with extended ascii, it's TUI. And man, loved the "absolute" keyword in turbo pascal. Instant screen writes when writing to a 2 dimensional array.
I don't remember "absolute" but I sure do remember "gotoxy" and it rhymes with boxy, yeah, you won't convince me otherwise.
It's a TUI if it uses text to build those elements.
You can be effective and powerful in any kind of interface, Just like you can be ineffective and weak in any kind of interface. People like TUIs because they're cool, and work over SSH.
Yes. A TUI runs in a text session. A GUI runs in a graphics session. A terminal emulator emulates a text session in a graphics session - and allows you to run TUI/CLI tools. This is apparently controversial?
> TUI runs in a text session. A GUI runs in a graphics session
What do you mean by this? I have never heard these terms before. I can launch and interact with a GUI from a text application, or a text application from a GUI.
It makes far more sense in the context of effectively a remote session or dumb terminal serial connection. Your "terminal" application is only emulating a text mode environment inside a gui, typically... you can ctrl+alt+F2 - F5 in a lot of Liknux's to switch to a terminal session if you want, but that's not what people tend to actually use.
Beyond this, without remote X properly configured, again, most don't and probably shouldn't.. you aren't running remote gui applications over an SSH session. Richer TUIs were pretty common in ye old days of DOS and other OSes before rich GUIs become the norm. DOSShell, Edit.com, etc. The IDEs of days past and Word Perfect even. These all interacted with Mice and were considered the norm. The features that allow this over a remote terminal today are pretty great IMO, the harder part is properly handling window sizes/resizes, etc.
With graphical extensions, there are even nice app explorers with image previews via TUI. It pushes the boundaries. For that matter, I often wonder what could have come with RIPscrip/RIPTerm if the leap to web didn't happen the way it did...
I think the single hardest part of TUI is dealing with wide characters and secondary fonts for color emojii that don't quite render in 2 spaces completely in a lot of termianls... it makes the line drawing harder too.
In the Linux world, your GUI is built on a stack of separate programs. At the bottom is the "display server", such as X11 or Wayland. One of these needs to run to have GUI windows (or a full-screen background) at all; otherwise the screen is just covered by 80x24 terminal, and even if there were mouse support there would be nowhere to click or drag to change that. Without the display server, you are in a "text session". It's relatively rare to do this locally nowadays, but Linux still adheres to the highly modular UNIX philosophy.
Your terminal windows (whether that's "Terminal" or "cmd.exe" or anything else) are still fundamentally graphical programs that emulate such a text session.
Is it character mapped, designed to run in a tty terminal? TUI.
Is it pixel or vector mapped, designed to run in a graphics terminal? GUI.
Of course strictly speaking TUI is a subset of possible graphical user interfaces, but the term GUI was coined to denote interfaces other than the already-ubiquitous text terminal interfaces.
TUIs have since absorbed GUI interface elements like buttons, checkboxes, and even pointer input, which I think is causing the terminology complaint here. Classical TUIs like Norton Commander are more about keyboard input and navigation. But being text-mapped is the identifying feature of a TUI, I think most people accept.
Sometimes your system doesn’t have a graphical session, like a raspberry pi with no x server running, or a cloud compute instance I’m ssh’ed into, or a docker image running on my laptop. Sometimes your system doesn’t have a (particularly usable) text system, like a work computer that disables the terminal or a family member’s MacBook who doesn’t have the time or space to install XCode terminal utilities to be able to use things like brew install.
My point is that it’s not a given that having one means you have the other.
TUIs are wonderful for the first case.
Perhaps he's thinking about "console" / "display server" but the lines blur fast (e.g. you can run GUI in linux console with framebuffer with some limitations)
- https://en.wikipedia.org/wiki/Linux_console - https://en.wikipedia.org/wiki/Windowing_system#Display_serve...
It feels like a reference to DOS graphics mode.
It's a TUI if it uses text to build those elements.
No. All you've done is make a low-resolution GUI.
TUI means "Terminal User Interface" or "Text User Interface"
A GUI that is built with Text, and intended to be used in a Terminal, is what a TUI is, colloquially AND definitionally.
What do you think qualifies as a TUI?
This is a pointless semantic argument.
Of course you can use the primitives of TUI, especially with mouse support, to reproduce a large amount (if not all) of the standard GUI interaction paradigms.
But it's bizarre, and missing the point from a UX perspective.
As an extreme example, we can imagine a program that displays the borders of a 40x15 "window" in the middle of a console, with box-drawing characters, putting a "close box" in an upper corner, with text like "File Edit Help" in the top left. We can imagine it responding to a click on the "File" text by popping out a "menu"; we can imagine a drag starting from the "title bar" causing the window position to be update (and the entire terminal window redrawn).
A lot of those kinds of functions, ironically enough, might make sense for a TUI editor implemented as a TUI (except the "windows" might just be understood as panels where the ultimate program displays parts of its output). But as an emulation of GUI windows, it'd be a strange, impractical novelty.
What's with the purism? It's just a term used to differentiate one way of making a UI from another. Who cares about what is practical when you're just trying to give a thing a name.
Even in your example, it's pretty clear cut. If the window is built with text and served in a terminal emulator, it's a TUI. If you build it with a graphical framework that now needs X11 or whatever, it's a GUI.
This is just needlessly pedantic.
I've been using TUIs since PR1MEOS mini-mainframes in the early 1980's, I know what I'm talking about.
The issue is not the text. It's the WIMP interface.
Define TUI. I'm genuinely curious what it means in your context.
Yeah, that's the point. Why did you think you needed to say it?
It's a GUI that works over SSH. There is a very valid use case for that.
I assume his point is that making stuff that assumes a mouse makes for a bad text-based UI. Absolutely fine if everything is controllable via the keyboard, e.g. if the tabs were labelled F1-Fn and they function keys switched them, or they had an underlined letter and Ctrl+letter switched focus to it, or whatever.
But if this thing requires you to just tab a lot through lots of pointless and rarely used fields to get to a "button" so you can activate it, because it's really all designed to be used with a mouse, then it's a bad text-based UI.
There are some incredibly good text-based UIs around, some going back to mainframe stuff from the 70s. Most of them are optimised for speed of control via keyboard rather than for looking pretty. Almost none of them would be quicker to use with a mouse.
absolutely, but it's still a TUI. Just like if you made a GUI that didn't have any mouse support and needed all keyboard shortcuts, it wouldn't stop being a GUI.
TBH I've always hated the backronym TUI anyway, so if TUI includes things that require mouse input, then maybe we can go back to using "text-based UI" for things that only require a keyboard like we used to 30 years ago.
Would you say old DOS applications like Borland's Turbo series of compilers were not TUIs? They ran in the console but had menus, mouse support, dialog boxes, etc...
How about those text games that used ASCII art and you typed in commands like "look" and "go north"?
I would say using text mode is the primary requirement for a TUI. The other requirement being some kind of human-machine connection, IE a User Interface.
> there is nothing textual about the UIs being shown here.
Well, except:
> a 1:1 representation of the concept within character cells.
TUI is build from text, and living within its constraints and what it's engine (usually the terminal) allows. GUI is build from graphics, and has basically a pixel perfect control of its own. This is a very notable difference, especially at the time when these terms were coined.
> TUIs are generally built for effectiveness and power
No, this is a result of different architectures and their constraints.
> But once you start adding mouse clickable tabs, buttons, checkboxes etc. you
TUI and mouse are predating the GUI (more or less). We had them already 40-50 years ago at the dawn of interfaces. We are now just moving back to them for practical reasons.
The UIs are text only, so they are textual. Modern TUIs may support mouse events. That this tool can export to several TUI frameworks is evidence that these UIs are indeed TUIs, even if not the most traditional.
“Modern TUIs may support mouse events” hah! They already did in the 80s…
Yeah, remember good old Norton Commander (https://en.wikipedia.org/wiki/Norton_Commander) - not that newfangled clone Midnight Commander, the original? Or the Borland text mode IDEs (https://ilyabirman.net/meanwhile/all/ui-museum-turbo-pascal-...), complete with windows, menus, buttons, scroll bars, file dialogs etc. etc.?
Oh boy, that sent me down the memory lane hard. I also had to Google “Volkov Commander”, as that name sprung to mind immediately too.
> That this tool can export to several TUI frameworks
It clearly cannot. Have you even tested it?
This? Alpha notice: Code export is not functional yet. We're actively working on it — check back soon.
Exactly, don't know why people are acting like actually makes TUIs, it's just a rough mockup of a TUI for now, with a convoluted figma-like UI.
I guess the headline and website was enough to get all these upvotes. Quite disappointing as someone in the early stages of making a TUI tutorial myself.
I've been juggling some BBS related projects myself that involve some TUI work over raw and web sockets that I've been working on... It's definitely a fascinating space and there's been a lot of relatively recent activity in the space.
> and there's been a lot of relatively recent activity in the space.
100 percent agree. I personally love what the openTUI folks have been up to. As weird as this might be to say, we're still in the early, early stage of TUI adoption.
No, a text-based UI is not sufficient. It must also work in a text-only session e.g., on the CLI over SSH.
Do UIs exported from this tool not worknon CLI over SSH?
I think your comment is nonsensical.
Zellij among is a great example, I can do everything with my keyboard, but every now and them I'm already with the mouse and just click a tab or pane, no functionality lost, just added, why the need to make a cutoff philosophical/semantic hard argument?
Good insight, but if you discount the visual elements (tabs, buttons, etc), you're limiting TUI to CLI, and I think that's unwarranted. The value proposition of both TUI and GUI is two-fold: you see the available action options, and you see the effect of your actions. So, yes, TUI and GUI _are_ closely related: who cares whether we're displaying pixels or character blocks.
Unfortunately, they are often artificially differentiated by the style of the UX interaction: TUIs promote the keyboard actions, and GUIs prefer mouse without corresponding keyboard shortcuts. Unfortunately for GUIs, their designers are often so enamored with WIMP that they omit the keyboard shortcuts or make them awkward. I hate it when, even if the ACTION button is available by keyboard traversal at all, it requires some unknown number of widget traversals instead of being one tab away.
Since the keyboard is almost always used for the textual data, it makes sense to me to always enable it for command execution. Well designed GUIs and TUIs provide both WIMP and keyboard UX, which sadly is not the norm today, so here's my vote to make them larp for each other more.
Would you make the same argument for classic UIs created with things like Borland's Turbo Vision framework? It's generally known as a TUI framework (including by Wikipedia).
I like TUIs keyboard-centric. Mouse can be a plus, but it should never be necessary.
That's fair... I feel that way about GUIs too in general though. Everything should be keyboard navigable and reasonable control flows. Tab and arrows, etc. Should be able to control focus and selection (enter).
I admit I don't always pay the most attention to it, as the UI components I tend to use do a good enough job of this. But I'm usually pretty consistent with it.
One justification for TUIs is remote access over SSH.
You can tunnel a port over SSH and get a web UI locally, though it's not commonly done. I feel like more people would actually do this if tunneling a port was just ever so slightly easier (like, you're already SSH'd into a box, then you run a command, then you somehow automatically get a tunnel for that command's UI port plus a local browser window open to the page)
While in an SSH session, press enter, then type tilde and capital C (enter ~C) and you can add command line options to the current session. To add a port forward from your local 8080 to the remote port 80 without closing the connection, do:
Thanks. This could really benefit from a TUI!
That is a neat trick. Added to the list.
(Ultimately unhelpful though because I use mosh everywhere these days and that doesn't appear to have anything fancy like this.)
Maybe it's just still too early in the morning yet, but what is the significance of hitting enter first?
SSH expects the escape sequence (tilde) to be the first character on a new line; since backspace is sent as a character, you can't just backspace over something you've started typing and then press tilde to have it recognized.
Technically, you don't have to press enter if you've not typed anything (try it in a new SSH session - as soon as you are logged in, type ~? to get the SSH help output), but since the comment was about doing this during an active session without ending it, I figured noting that pressing enter first to be sure you're on a new line wouldn't hurt
I like TUIs because I run everything in tmux and I can just pick up work from wherever I was on any computer, phone or tablet.
share some good (easy on remembering keyboard & mouse) tmux configs. I usually struggle with copy pasting many scrollback lines from/to tmux. would love for my claude to be natively tmux aware.
I do this a lot but I'd still prefer TUI where possible. With too much visual content it isn't of course, but for many cases a TUI is much more responsive and much lower resource.
I largely agree with you, but there are limits to what a tui can do well. If analyzing a flame graph or performance trace, web UI is a better fit. However, most things are not that.
Even easier is just using an X server, if you have it set up properly you just need to run the remote app and the window pops up on your machine.
(I think terminal-based GUIs are neat just for fluidity of use- you can pop one open during a terminal session and close it without switching to mouse or shifting your attention away from the terminal. They can also be a nice addon to a primarily CLI utility without introducing big dependencies)
Yeah I love that about X. I remember in the 90s when I first figured that out. I was logged in from a university workstation into my home computer with SSH and I launched my mail client or something and I thought doh, stupid that will only popup locally.
Then colour my suprise when it popped up on my screen right there. Slow as molasses but still. Wow. Magic.
It's a shame Wayland dropped this. Yes I know there's waypipe but it's not the same.
> It's a shame Wayland dropped this.
It... really isn't. Like you said, remote X was barely usable even over an entirely local network. Most applications these days are also not designed for it, using loads of bitmap graphics instead of efficient, low-level primitives. So you end up being just one tiny step away from simply streaming a video of your windows. We have better tools for doing things remotely these days, there's a reason approximately no one has used remote X after the mid-90s. It's a neat party trick, but I don't blame the Wayland authors for not wanting to support it.
It is, there were tools like NX that made it entirely usable even latencywise. And these days we're really going more and more to remote computing.
In the time when wayland was invented it made sense because we did everything purely local. But now it's as outdated as X11 was in 2010.
And yes I still use it a lot. It works well. Networks have become a lot better and even most cloud compute I use is geographically nearby.
What made it slow back then was that I only had a 128kbit uplink at home. And the uni had 2 mbit for the whole computer science building :)
> In the time when wayland was invented it made sense because we did everything purely local.
People complained of no forwarding in Wayland when it was invented.
> one tiny step away from simply streaming a video of your windows
In the 80s/90s this wasn't feasible due to network latency and bandwidth, but it's pretty common now to do exactly this, with VNC and other remote desktop protocols.
Like what? X forwarding has pretty much always been the thing most likely to work for me and I haven't been able to find any equivalent.
The big obvious one is web-based tooling. Your information & settings are stored on a server and you use a web browser to view it via whatever device you're on. For more locally based workflows, we have networked filesystem protocols, automatic syncing between systems, that kind of thing. It's not a 1-1 equivalent of running a remote program and viewing it locally obviously, but it gets the same job done, in a much more useful & flexible manner than X forwarding did.
For example, the remote mail client usecase I was replying to is simply done with a webmail client today.
I don't really feel like web interfaces or syncing are really a substitute tbh, and I'm not sure how they're more flexible. ssh -> run -> gui opens, and the program itself doesn't need to be designed differently to work
> and I'm not sure how they're more flexible. ssh -> run -> gui opens
But this doesn't work on your phone, or on a Windows or macOS device, right? That's what I meant by flexible, X forwarding fits a pretty narrow set of usecases, while on the other hand keeping programs on the clients and data centrally located on a server allows for a whole lot more options for how to interface with that data.
(To be clear, nothing wrong with X forwarding! It's a cool tech and I'm glad you have a use for it! I'm just arguing that it's fine for Wayland to not try to support that kind of thing, because we've got other ways of working remotely now.)
X servers are available for phones, Windows, and macOS. X interfaces not designed for phones can be difficult to use on phones. But web interfaces not designed for phones can be difficult to use on phones.
There is not a web tool for every use. And web tools are not better for every use.
IIRC, it's not that secure though.. I'm really surprised people didn't do more things like send animated skulls to people's desktops.
Ps: oh yes and before '93 I've had so much fun practical joking around :)
Xauth fixed that way back in '93. All you have to do is use -Y not -X with SSH.
Waypipe looks interesting.
The main advantage of x forwarding for me was when I'd randomly need it and had nothing set up ahead of time. Hopefully it starts getting installed in distros by default eventually.
I'd rather use a TUI than a web UI.
Sure, but my point was that UX matters for TUIs. A TUI with a UX that fits its paradigm , again like lazygit, works great over SSH.
Another justification could be simply some people like using them better.
People don't build TUIs because they want to run apps in the terminal, they build them because the terminal happens to be the most portable app platform available.
I've been working with notcurses recently and it is a full TUI that handles mouse events just fine. Runs over slow SSH connections and everything. The nice part is that you can fully operate applications built on top of it with the keyboard if you so choose, the mouse is just a shortcut.
Sadly the project is not really in a usable state at the moment. The documentation is incomplete riddled with errors, the code has some pretty glaring bugs, and it's close to abandoned. It's a shame because you can do some really amazing stuff with it.
You might not like this type of interface, but it is hardly "nonsensical". In the 1990s this sort of text-based GUI was common in DOS programs, such as Borland's "Turbo" languages and the original pre-Windows FoxPro.
lazygit supports vim style keybindings and mouse click and scroll. I mostly use the key shortcuts but sometimes the mouse is useful. But i agree that a well thought out state machine that can be navigated through via keyboard is a dream to work with. Lazygit is superb. But this is not a distinction between TUI and GUI.
Yes and no. Early DOS UIs had elements of TUIs and GUIs, and supported mice. Many old school greenscreen applications were like this too.
> But once you start adding mouse clickable tabs, buttons, checkboxes etc. you left the UX for TUIs behind and applied the UX expected for GUIs, it has become a GUI larping as a TUI.
TIL that VIM is not cease being TUI the moment I type :set mouse=a.
Hot taking, LARPing and teenage angst (caused by generational gap with those has been using TUI since 1980s) is on your side.
My ancient boxed copy of Visual Basic for DOS 1.0 that supported mouse clicks on TUI buttons would have found your viewpoint quite offensive if it had any AI in it ;-) Oh boy, good old days.
The distinction is - if it runs over ssh (no x / graphics login) or on a headless machine - TUI
If it requires graphics login, even if it uses character layouts - GUI
IMHO the T/G is not for the display elements, it's for the type of session.
Not to put too fine a point on it, but X11 runs over ssh just fine. No "graphics login" required.
Man, I've had so much frustrating just trying to copy & paste from inside a terminal running e.g. opencode or crush.
I think TUIs are neat, I guess. But I think these things have abused the concept extensively. They don't actually interact well with the rest of a Unix environment.
Hold down a key (alt, I think) to prevent clicks.
This is exactly the kind of passive aggressive attitude that is tolerated on HN that makes this place unbearable.
"This is dumb" - gets downvoted to oblivion. "This is nonsensical + a bunch of absolutely bs reasoning" - second most upvoted comment atm.
HN tolerates the appearance of quality discourse over the actual thing, and dealing with this dissonance in most comment sections is exhausting.
Reddit moment!
As a german, I say:
UIUIUI
Drawing a “nonsense” line between TUIs and GUIs is pretty arbitrary, it’s all pixels on a screen at the end of the day. People like the TUI vibe, and that’s a good enough reason to make and use them.
I love TUIs but one main reason for that is that they're keyboard centric. If I have to use the mouse it kills it for me, if both work then it's fine. I hope that modern TUI makers keep this in mind. What's great about the keyboard centric is that with a few keystrokes/shortcuts it's very easy to do repeatable work and takes less energy than hunting boxes to click on with the mouse.
TUIs aren't more inherently keyboard driven than well constructed GUIs. You can easily make a keyboard driven GUI that has all the shortcuts you'd add to a TUI. (Just don't let the "UX design experts" near it.)
I actually agree with that. And I enjoy the fact that TUIs are becoming popular. But there is more to it than just the 'vibe'.
The vibe might not be a necessary reason, but it is a sufficient one.
I really don't want my TUI's to look like GUI's rendered in low res. The appeal to me of a TUI is that it is built specifically to be a TUI, and that means eschewing complexity and detail, and favouring compact text.
> GUI's rendered in low res
That's literally what TUI's looked like starting from the late 1980s and throughout the 1990s... You have a pointing device, might as well make use of it to enhance discoverability.
This seems really reductive. Some UI paradigms work better at 80x24 vs at 640x480 (never mind whatever resolutions we have access to today). Or rather, the 80x24 text grid is using more pixels than that, but everything is aligned to that lower resolution, and that fundamentally changes what makes sense to do. Floating windows that can be dragged around to arbitrary positions? Terrible for low-res; classic for higher res. Dividing lines that split the screen into panels, and can be moved around a row or column at a time with a keyboard shortcut? Pretty much the opposite (enthusiasts of "tiling WMs" might disagree).
Didn't they evolve from that because better graphics was better? Otherwise why not stay text if there is a huge advantage in all Text made graphics?
This is why I don't like TUIs at all, they're really bad at displaying complex information, handling complex interactions, and discovering how to compose those together.
Interesting idea, but:
> Design once, generate production-ready code for your framework of choice. Switch targets without touching your design. Alpha notice: Code export is not functional yet. We're actively working on it — check back soon.
In other words, it isn't at all usable right now. You can't produce a TUI with it, not even a limited one.
You can still design the layout. That’s useful, but not nearly as useful as they are planning.
Why do you believe anything the site claims? It might all be hallucinations anyways, and others report not even being able to open the app.
The app certainly opened for me. I don’t know that I’d use it. Design and layout of an interface is useful apart from the implementation though.
Ignore the haters. This is an excellent idea, I'm getting some old Borland vibes. Keep it up, can't wait to see where it goes!
Ignore the haters, sure. But don't ignore the well argumented criticism that you're getting from an overwhelming majority of your peers, as it's happening right now.
>can't wait to see where it goes!
Fall into this toxic positivity nonsense at your own peril.
Toxic positivity? At your own peril? OMFG.
Let's imagine one do. What do you think can actually happen that is so negative? Toxic TUI will hunt you in dreams?
It doesn’t even exist, you’re getting excited for a hallucination.
"Majority of peers" has never experienced proper TUI in their life, and their opinion is hardly relevant.
Vibe-coded trash, even says so in the Readme. Not sure why this gets voted to the frontpage.
Better get ready for almost all software to use AI assistance in its creation.
You can build great things using AI agents, and you can build trash.
Your ideological opposition to that is not shared by as wide a percentage of developers as you may think based on some highly self selected online corners.
Way to read something into what I have never even wrote and try to spin it as something ideological. This is not "AI-assisted", it's completely vibe-coded by AI and the software doesn't even make sense since you can't export anything. It's just low-quality trash dumped on Hacker News and I'd argue this is not the place for it.
It's not anything though. It's a website and electron app that promises functionality that completely isn't there. It's useless, but instead of being art, it promises functionality, so it's functionally trash.
I too remember running `rails new MyGreatApp` and having hoop dreams of being the next billionaire entrepreneur, but a boilerplate app is a boilerplate app.
Of course you can build great things with AI, but trash written by AI is worse than trash written by a human, and some things are just trash.
theres a difference between AI-assistance and vibe coding. One of them requires you to know what you're doing and make good design choices
If 30% of devs love openclaw, I'm not re-examining my informed opinion, I am forming a new one about that 30% cohort.
That was a reaction to "Not sure why this gets voted to the frontpage.", not to any correctness claims.
Apparently we now write desktop applications intended for designing the UI of other desktop applications, in TypeScript that runs in a Docker container, using a bunch of web frameworks (including for CSS) and self-hosting an nginx server?
I would have expected a TUI editor to be itself a TUI.
Funny how you can tell a project is vibe-coded just from a first glance at its website. All these websites seem to somehow have the same visual style. Anyone noticed this?
Yep, my fans started revving as soon as I loaded it. Animations are out of control on the normal web as it is, but genAI sites take it to another level.
This is why I recommend NoScript.
Same, and it lags simply trying to scroll down the page. Unacceptable.
That, too, though I'm sure that particular problem is mainly because of the textual animation in the background.
Yeah it was. I've been hoping to get the time to write a userscript to identify and stop animations like that, but I haven't been able to understand the whole requestAnimationFrame thing enough yet.
What is the point of having this if code generation is not functional yet? That is the entire point of this app.
TUIs are very in right now. Nostalgia/vintage-computin-aesthetics & guru-gatekeeping around command line savviness are front and centre in the HN-and-adjacent mindshare.
It's strange. For decades we've been trying to move towards GUIs, now we're moving backwards.
To show off an AI generated website
Exactly.
This is going to end up with TUIs that resemble old BBS ANSI art, such as https://16colo.rs/
It completely misses the reason people like current TUIs.
I agree. The animation on the site lost me when it placed a button. IMO, buttons are not part of TUIs. Those are just low-resolution GUIs, IMO, and that’s sort of the worst of all worlds. The first good TUIs were things like top and elm.
FWIW, I still love to see the old BBS UIs and ANSI art. But that's probably just nostalgia talking.
FYI LLMs are great at generating the ascii art, so you can create real fun games and TUIs that look like old school BBSs.
We can remain grateful the kids haven't discovered how to use figlet in HN comments.
ENSHT comes for everyone. This is sexual selection over natural selection. Claude Code also gets this wrong, they got way to fancy and ruined what a good tui is by being an uncanny combo between a scrolling log and a completely rewritten canvas.
In this age, rich TUI's feels wrong to me. Tools that expose a minimal web server with a lightweight UI is much more welcome than a complex TUI. But for most interactive terminal apps, it feels more natural when there is a single input at a time, like a wizard interface.
This looks really cool. However, the current AI models are pretty good at designing UIs from prompts and even turning screenshots of mocks into full UIs. I'm not sure this visual design approach would save time vs simply prompting an AI agent.
That being said, I could see a niche market for a designer persona who is used to building in tools like figma.
also, this assumes humans are still the primary CLI consumers. With agents increasingly being the first-class users of command-line tools, building visual design tooling for terminal UIs feels like optimizing for a shrinking audience.
Well you still need some human prompting the AI and looking at results, no? :)
Agents aren't picky with UI, so most effort will always be spent designing for humans, even if they are not the primary consumers.
If you want inspiration on all kinds of TUIs on show and display Terminal Trove [1] is useful to get an idea for what other tools look like.
I find the search [2] also helpful.
[1] https://terminaltrove.com/
[2] https://terminaltrove.com/explore/
For exports, it is missing the ultimate: Borland Turbo Vision, the Rolls-Royce of TUI frameworks.
Why are these things being built on web technologies? There's loads of "modern" terminals that use typescript etc. to me terminal means lower level.
Also wheres the Linux version? You've Mac, windows, and docker. When someone says terminal to me I default to Linux.
On one hand this is a neat idea. I've thought about how nice it would be to have a visual layout tool for text-based designs. The current offerings are slim. Of course, you could easily argue that if you need a visual tool for it, you've gone too far; even the most sophisticated TUIs are still extremely simple.
On the other hand, for this work as they describe, it needs to be a complete UI framework across a bunch of languages and built on top of a bunch of existing frameworks. That seems... ambitious. Building one UI framework for one language is plenty hard enough.
Watched the video. Why isn't the editor a TUI itself?
Because a website is easier to use and more accessible.....
This one is not very accessible, try using tab + arrow keys to focus anything on the sidebar.
Great question.
The TUI hype seems like nostalgia for COBOL mainframe apps that most people have never even used. A sort of secondhand cyberpunk role play with zero focus on actual UX.
Also if TUIs are so great, why isn't this a TUI app?
Have you ever watched someone USE those COBOL TUIs? Everyone from airline ticket agents, to local governments, to folks at Home Depot while looking up inventory. They could fly through menus and accomplish things. I remember when Best Buy switched to a Windows-based experience. It was terrible. Simply adding a mouse+windowing experience slowed everything way down. I saw it first hand at Target too. They went from an OS/2-based TUI to Windows NT. I know there'll always be those folks that think we're all just trying to play "leet Haxorz", but there's just something about those systems that people deeply connected with.
I personally think the opencode and kilo CLI have great UX, certainly better (and easier to use) than the web versions of both.
A lot of the recent TUI apps are really not old-school in any way. Not all apps need the feature-set of a browser engine. And compared to native mac/linux desktop apps, TUIs get cross-platform support by default.
> Also if TUIs are so great, why isn't this a TUI app?
We all know the answer to this
TUIs are great coz they work seamlessly over shell, but there is no reason for that for editor.
Half surprised there's no raratui export with the other options. That said, probably lends itself more to Ink and @opentui/react. Also slightly disappointed at the lack of a direct Linux build for AppImage and/or Flathub. Also not using Github's releases which is a little curious.
This is a vibe coded app and isn't what I'd want but still, it's interesting to consider what a good implementation of "Figma for TUIs" could be, especially if it avoids the trap of simply treating the console as a crude raster instead of taking advantage of text and keyboards. IMO we don't need WIMP GUI shoved into terminal emulators.
The lack of accessibility of TUIs is not great in general.
I'd much rather terminals emulator provide a webview directly, and maybe use https://webtui.ironclad.sh/ if you really want the look.
I think it makes more sense for a cli to offer a mini webserver instead.
Think `fish_config`, but opened in the terminal directly [0].
[0]: like https://iterm2.com/browser-plugin.html
> The lack of accessibility of TUIs is not great in general.
Interesting. In what ways? I haven't heard anyone express this concern before.
Depending on how the TUI is made, it can be very visual, but lacks structure for a screen reader (unless you stay in the very simple "input field: value" kind of prompt, but even then auto completion is tricky).
Web browsers offer the DOM to tools such as screen readers (OSs offer their own accessibility sdks). Someday perhaps the TUI application could talk to the terminal emulator that would itself talk to the accessibility sdk of the OS and that info would somehow then be accessible.
There was a beginning of discussion at bubble tea[0] about this for example.
[0]: https://github.com/charmbracelet/bubbletea/issues/780
TUIs seem like they should be well-suited to screen readers, but yeah accessibility is often unfortunately far down the list on average project priorities.
TUIs sadly have become niche tech, and accessibility is rarely a concern for niche things.
A UI design tool for TUIs -- made with Electron?... fun times!
I give it a month before someone launches a TUI-TUI.
You can run it as a web app, no need for electron.
Just `bun run dev`
That is concerning.
Look up Visual Basic for Dos for a surprisingly good TUI editor!
Old VB is still king for laying out interfaces, for the narrowest of use cases today, bu still... Makes me sad how much we regressed due to the web.
Yes!
So we’re going full circle here right? Can’t wait for the first TUI MVC/MVVM/MVP/M-whatever framework to show up.
The TUI-vs-GUI debate in the comments is probably not the most useful frame. The practical question for something like this is whether the exported code models terminal layout constraints correctly. Textual and Ratatui both use constraint-based layout, not absolute positioning, and a designer that works in pixel coordinates will produce UIs that look right at one terminal size and fall apart at others. Has anyone tested an export against real usage?
Nope, check out something like wiretext, look at this example I put together very quickly
https://wiretext.app/w/WUtjS1bk
TUIs built today should be usable by AI agents. I'm not sure exactly what it looks like but I'm imagining that every UI view has an associated CLI command that can yield precisely that view. Maybe like formally structured breadcrumbs, or maybe like Emacs "keyboard macros".
I've found agents effective using GUI apps with nothing but the ability to take screenshots and send mouse and keyboard commands. I imagine they'd work even better with a TUI, even if it's not designed with agents in mind at all.
Agents excel at using CLI tools with well-written "--help". So maybe consider that instead of TUI.
Yes, they do, but the premise in my comment (and this discussion) is that a TUI is being written today.
This is so cool I immediately wanted to convert my apps. But then when I thought about it, well it's trying to recreate CSS but in a majorly worse way.
Browsers are ubiquitous and I can just tell ai to build a web page. I can't really see a use case other than novelty.
If you need a UI over SSH or inside tmux, skipping browsers and CSS isn't just a novelty, it's essential since HTML can't touch that territory. TS-based layout in terminals can be ugly but it also dodges a pile of accessibility, latency and bloat issues you get by default with anything running in Chrome.
k9s, ncdu, htop, powertop are good showcases how a TUI reduces mental load and are superior to browsers and / or other GUI tools
More importantly, it also reduces CPU and memory load.
Somewhat related: Tachikoma.jl can do windows inside a terminal UI. https://discourse.julialang.org/t/ann-tachikoma-jl-a-termina...
The biggest pain point with TUIs has always been the design iteration loop. You're basically writing code blind, running it, tweaking numbers, running again. It's like writing CSS without a browser preview.
Something like this could genuinely help for the layout/positioning phase, even if you still hand-write the interaction logic. The debate about whether these are "real TUIs" kind of misses the point imo. Textual and Ratatui already blur that line with mouse support and rich widgets. The ship sailed on pure keyboard-only text interfaces a while ago.
What I'd actually want from a tool like this is to export to multiple TUI frameworks. Right now you're locked into one ecosystem and the code export isn't even working yet, which makes the whole thing feel premature.
A TUI library with hot reload would be pretty cool
Turbo Vision and Clipper want their glory MS-DOS days back.
Nowadays we have Unicode characters and better colors though.
MS-DOS always had better colors than UNIX, a framebuffer isn't the same a vt100.
I do agree Unicode is better than code pages, or doing alt + num pad codes.
And Ashton-Tate's Framework IV on an 8088 with a MDA display.
Have been spending so much time in the terminal lately for the first time in my life (non-developer here) that made this for fun to spruce it up: https://github.com/dvelton/terminal-profile-studio
I built something like this in 1993, it was used to design layouts for DOS apps and headers for printed listings. Imitating the BorlandPascal and Novel TUIs of the day
I'm not sure the utility of this kind of stuff anymore. It's relatively easy to sketch a layout on a napkin + prompt and then prompt claude code to use python textual as as TUI layer. I've had pretty good success with Textual+Claude so have a few colleagues. You could probably use Figma + claude etc as well.
The hardest part of TUI development has always been the layout model, not the widgets. Most people start with absolute positioning and then discover their terminal is 80x24 on the CI server. A visual designer is useful if it outputs layout constraints rather than fixed coordinates. Otherwise you end up with UIs that only work at one terminal size.
Missing Ratatui[1] support.
[1] https://ratatui.rs/
> No install fuss — download and start designing immediately.
also
> Gatekeeper blocks the app immediately. You'll see either "TUIStudio cannot be opened because it is from an unidentified developer" or "TUIStudio is damaged and can't be opened" on newer macOS after quarantine flags the binary. To get past it: right-click the .app → Open → Open anyway — or go to System Settings → Privacy & Security → "Open Anyway".
tbf that's Apple's fault, not the choice of the free, unpaid open source developer.
Apple's fault that they didn't bother to edit the text that says "No install fuss"?
Probably don't know how now that the LLM helping them write the code has lost that context.
From their github it appears all the code is llm-generated
You mean the AI agent that was prompted to vibe code this?
For {root} sake I'm a designer. Mostly all the code has been written by Claude and ad latere.
Probably why the actual product, the code export, isn't working. I doubt it ever will. Neat toy, though.
Guess who's back, back again?
VB's back, tell a friend.
The one thing these always miss is image protocols. Do you plan to support terminal image protocols like sixel, kitty image protocol, etc.?
tip: your git repo's description (not readme, repo description) does not link the website. It should.
Also fill the Website field in About section.
Gotta say I did sort of expect this to be a TUI app itself.
This looks really cool. Is the use case of getting an LLM to respond with custom TUIs something you have thoughts about?
That's cool. I literally vibed something similar a month ago for myself!
I wonder if one of the LLMs could generate code from a screenshot of a layout designed by this.
Claude Code built a TUI for me last night, in this case to step through nanosecond timestamped ITCH market data messages and rebuild an order book visual in the terminal. This type of stuff would have taken a day - but done in 5 minutes now.
There's something incredibly ironic about a visual tool for designing TUIs...
We got a RAD IDE for terminals before GTA6 and before anyone sensibly makes a replacement for Electron. Wild.
This is really cool though.
I seem to remember having a RAD IDE for terminals with Turbo Pascal back in the 1990s. But yea, still before GTA6.
I would be REALLY REALLY impressed if it manages to do this without bugs. Just using pythons textual can be very complex, belive it or not. Maaging not only to that but other frameworks too sounds insanely complex. I have a strong feeling this is vibecoded from the commit history?`
Ah yes, it says clearly that on the github page. Still, if its works, I am then impressed by the LLM.
Edit: It does, in fact, NOT work for code export. Level of impressiveness massively dropped.
Yeah, the website has many bugs too. Literally can't click on 50% of the "clickable" stuff. Not impressed by vibe coded nonsense. The comments here are weird, people are discussing the "idea" rather than the broken implementation.
Probably a bad omen of things to come for the internet.
The irony that a TUI studio is not written as a TUI...
Am I the only one who thinks the recent TUI explosion is absolutely not necessary?
I mean yes, code editor are great for this but a lot of the TUIs I see are so slow it begs the question why they exist to begin. CLIs are supposed to be remixable and scriptable.
I think a better architecture would be to generally keep CLIs work like CLIs and have separate processes that add terminal rendering functionalities for those that need / want it but in general it is an anti-pattern to start from this as default.
I feel like they are a workaround to GUIs being slow and bloated Electron apps.
But I wish we'd just make fast GUIs instead of giving up and building TUIs instead.
Depends highly on the specific application. Take a simple example of looking at process usage. You can use ps from the command line to get all sorts of info about a process. But there’s no substitute for top to show you an updating list of top cpu consumers, which ps just can’t do.
That's roughly aligned with my thinking. Make it a CLI. And if there's a lot of configuration that you can pass to it, have an option for rendering those options as a TUI.
Of course it's not necessary, it's a fashion. Choosing to make a TUI instead of a GUI is a fashion statement, it signals aesthetic alignment with nerdy shit and says the program isn't meant for common proles. There's pretty much nothing a TUI can do that a GUI can't do, while the opposite is very much not the case.
See also all the programmer blogs that feel the need to use a monospaced font for prose, to signal that they're a programmer.
this is a cool idea lol but is a pretty nonsensical explanation of what you can even do with it
The background ASCII animation is so cool! Is it an actual simulation?
Use the source Luke! It's an "ASCII plasma background" rendered into a canvas element.
Turbo Vision strikes back
One can only dream
Missing ncurses support.
So this is a TUI WYSIWYG GUI ?
I wish HN had flairs (tags) like Reddit and mandated a few for AI-related work (AI-Assisted, AI-meta, AI-vibecoding) or something so these could be filtered out
Seems nice.
I launched https://github.com/alganet/tuish yesterday (pure shell backend).
Exporting to pure shell could be a killer feature, especially for smaller and ad-hoc apps (no dependencies, no compilation, etc).
The website UI is unreal, I loved the idea
When your TUI is so complex that you need a GUI to design it, perhaps you shouldn't use TUI in the first place.
I'm not sure that's a fair criticism. Many things require or benefit from something even more complex to make them (car -> factory, code -> IDE, text -> editor, food -> kitchen). I think the real debate here is that which is found in the other comments: do we want TUIs to look like GUIs?
Why did they make a website?
The corners of the boxes appear in the wrong place in the cell.
I don't think there is utf8 characters that allow for drawing on the outside of the cell, (happy to be wrong)
┌ (U+250C), ┐ (U+2510), └ (U+2514), ┘ (U+2518) <-- these 4 draw in the middle of the cell.
｢ (U+FF62), ⌟, (U+231F), <-- these are two that cover part of the outside, but not the other corners.
｢┐└｣
Can anyone tells me how to get those 'corner of cell' characters, including uprights and horizontals ?
I really hate these pointlessly dynamic website backgrounds that make mobile devices hot to touch. Unfortunately vibe-coded websites love these.
No ratatui?!
This is like QTdesigner but for the terminal. Huh.
Noice figma for terminals! Dude super cool idea, great job =D
I want something like that, but for Bootstrap,Tailwind or Quasar
Anyone notice the computer image at the top of the page doesn't have the right number of keys?
The fact that this isn't a TUI itself is a bit disappointing.
The fact that even the preview isn't a TUI is just lame. Keyboard controls are also non-functional right now.
I don’t want to be a curmudgeon, but why not just use CSS, HTML, React, etc. at this point? You could choose a style that looks like a TUI.
But will it render in a terminal over ssh?
(I know, I know, port forwarding should work for a web app.)
I find it slightly annoying and disappointing that the blocks saying what frameworks it's designed to export to aren't links to those frameworks.
this looks insanely cool.
One of the most original ideas I have seen on HackerNews in the past few years.
This website eats a whole CPU core
Another W from the web developers mafia
+1 on this
No idea why this is hyped up these days.
The only reason I can think of is what I said before [0] given that the web was destroyed by the same web developers, then so was the desktop (with Electron) and now of all places terminals are now getting destroyed with infinite slop like this.
This nonsense will continue and accelerate until it reaches hardware.
[0] https://news.ycombinator.com/item?id=47210187
Amazing cool design tool for TUI's I got it running instantly and it feels stable and complete as well. Only 10 stars in GitHub.
Ha, well proof that AI let's you build anything you can imagine. Wait till I show you Remote Desktop, one day macOS and Linux will catch up.