Linux desktop desperately needs:
1. a good raster photo editor with non-destructive editing capabilities
2. a good vector graphics app - Inkscape could be it, but it would need heavy work on stabilizing it, and UI improvements around anything involving mask (though to honest the bar is low here, Illustrator has horrible UI around masking to)
If anyone can do these things well, I'd suggest a crowdfunding campaign on Kickstarter or something. Lots of people will throw money at anyone who can produce something intuitive and stable! Not need for tons of features, just make it fast and stable, and someone will add on. And using a "hipster" new language like Rust would also help the project's popularity with new developers... few people are willing to hack on an open source project for the "fun" of improving their C/C++ skills nowadays...
Check out Scribus. I'd describe it as more of a serious publishing app, where Inkscape may be more for artists. Scribus uses less resources and supports multipage documents.
I've also had issues with the cage transform in Gimp, but it's fixed now, so if you're on Ubuntu, just add a PPA with the latest version of Gimp.
> 1. a good raster photo editor with non-destructive editing capabilities
I think Darktable does non-desctructive editing but I haven't used it myself.
> Gimp and Inkscape are both extremely sub-par apps... Paint.NET beats Gimp in usability, Photoshop in both features and stability.
I think it's a bit sad to be whinging at those good pieces of free software for lacking features, and just recommend proprietary software instead. Surely, it is also the developers of the proprietary software that are causing this lack of features by not sharing their code for those.
It also suggests you are willing to give up your freedom for a few extra features. What are those features missing in Gimp? (Gimp does have CMYK editing, if you install the gimp-plugin-registry package.)
> Linux desktop desperately needs:
Those non-free programs you mentioned can probably be run with Wine on GNU/Linux.
> give up your freedom for a few extra features
Not really, I use like 10% max of Photoshop's features I guess. But those features simply feel perfectly implemented, they just f work. Heck, I even use it for vector graphics sometimes because it "just works" and feels more natural than Illustrator.
And I'm not just used to PS. I actually liked Corel Photoshop more back in the day.
And Inkscape has the same natural/intuitive feel to it after you use it a bit. Starts to crack once the complexity grows, but for starter it just works. Blender has a learning curve, but afterwards it also feels "natural" and "well built".
Gimp is just bad at... everything: (1) performance feels horrible, like doing everything in slow mo compared to anything else (2) UI feels very inconsistent and has very weird concepts about the relationship between layers and selections and stuff like that (like "wtf is a 'floating selection' thingy doing in my layers panel? ...you get used to it, but boy it's weird), keyboard shortcuts don't work in all contexts (sometimes it feels like they randomly stop working) and (3) there's no easy way to reversibly do things like ad an adjustment/effect layer, turn effect on and of on a layer or some groups from the layers panel etc.
I don't care about color space support or even mildly advanced features. And I could put up with a "weird" UI if I could find some logic or intuition for it. But at least stellar performance and stability would be requirements. If you can do things fast enough and things don't crash at 30 huge layers, even non-destructive edits cease to be a "must have" since you can just clone stuff like crazy and always have multiple "versions" of the "same" layer in the same doc. That's how lots of professional graphic designers use PS anyway. I was once handed a 20GB "trashcan psd" from a designer, and I hated every second of touching it, but it miraculously worked, relevant content was extractable, and the job got done.
Honestly, its hard to say what's wrong with Gimp, because I can't figure out anything fully right about it. Even the fact that I'm comparing with Adobe Photoshop, a program itself known for horribly beginner-hostile and unintuitive (but in a weird way "learnable" and "natural") UI is a sign of how off things are.
Inkscape has also been crashing on me a lot in the past and was way too heavy, but since they released the 0.91 with the new Cairo-based renderer, it's been pretty fast and stable.
I've personally had a harder time with Blender. I've found it particularly confusing that about half of the functionality seems to just not work when using the game engine. And when importing a video for editing, the length of the video and the audio may not match up.
Ugh, this is so true. I've never done professional graphics work, but I've used Photoshop on and off for small projects since they added color in version 2.0. I WANT to like Gimp and I give it a shot every year or so, but it never clicks. I mentally prepare myself for a different experience and try to keep an open mind, all while considering that my comfort will change with use, and I always leave disappointed and frustrated.
I mean, how do they make simply selecting something difficult? Even layers were unintuitive last time I tried Gimp. I've used a lot of janky software, but Gimp is the most bewildering case. You immediately see really good ideas expressed in the UI and then you interact with it and the whole thing collapses.
I thought maybe I was remembering things wrong, so I fired Gimp up for the first time in months (v2.8.22 - I keep it updated with Chocolatey). It took 2 minutes to load (no image) from an NvME drive on a quad core system with a 1080Ti. That's 1.5 minutes longer than any other program I have installed. Probably not optimized at install-time, when the user has time to burn, and instead opts for optimization at first launch, when the user actually needs to use the software. Bad start.
Open image. Select part and adjust brightness. Click OK. Click and drag outside selection... and the brightness dialog pops up? But, wha... B/C is a tool. Weird but okay. Apply filter, click and drag, and B/C again? Undo, wait, no, have to cancel. Alright, let's move this selected area over... Why is the entire canvas being moved? Undo. Trying rotate... 3 icons from the move tool on the same line, but it only rotates the selection. Of course. Rotate selection and hit enter... nothing. Oh, the rotate dialog isn't in focus. No, wait. It is. Enter does nothing. Unfocused? Now Enter works.
Adding a layer. First click on new layer button not detected. Second works. Select background layer and marquee a small part. Ctrl+C. Select newly made layer. Ctrl+V. Probably can't see it because it's pasted on top of the exact same part of the image. Select move tool, click and drag, no, nothing was pasted. Click and drag again to be sure and the part I rotated earlier now moves across the screen. Hide background and 'floating selection' to confirm nothing pasted.
3.5 minutes and about 15 undos/cancel button clicks to do 'all' that. I repeated the above in Paint.NET, including launching, in 15 seconds and 2 undos from one minor quirk.
It's like a luxury car assembled inside out. In theory, you have a luxury car, but in practice you're sitting on cold metal with an engine block obstructing your view of the road. Sure, a bunch of mechanics could take pieces off and attach them properly, but you've been driving the car around town for 22 years and this is the best they've managed so far.
PC: 6700k OC @4.5GHz, 32GB/1TB NvME (4x PCIE mode), 1080Ti, most background processes off: GIMP 1st run: 90 s. GIMP 2nd run: 23 s. Paint.Net 1st run: 3 s. Paint.Net 2nd run: 2 s. Paint3D 1st run: 2 s. Paint3D 2nd run: 2 s. Krita 1st run: 7 s. (v3.3.3 first run ever) Krita 2nd run: 4 s. Gravit Designer: 6 s. (first run ever) Visual Studio 2017 takes 18 seconds with a handful of extensions.
Ugh, VS2017 update out...
Outdated version of Photoshop WHILE updating Visual Studio, 1st launch: 17 s. Krita WHILE crypto-mining on 3/4 logical processors (pinned to 100% usage): 4 s. Paint.NET while opening 30 tabs in Chrome: 4 s. GIMP while opening different 30 tabs in Chrome: 25 s. (reduce the chance of cache loads, not scientific, Chrome finished loading first, by about 10 seconds)
On a Surface Pro 3 (Core i7)... Krita while installing Windows updates, 1st run ever: 17 s. Krita 2nd run, updates still installing: 10 s GIMP, no updates installing, clean GIMP install, 1st run: 243 second. Yes, 243. GIMP 2nd run: 17 s.
Tried GIMP again on my desktop, no unusual background processes. 3rd launch: 18 s. 4th launch with mining pinning 3/4 CPU: 23 s.
So, overall, one of the most often recommended FOSS alternatives, even by ME, is a distant last place in these extremely ad hoc, unscientific, off the cuff tests across 2 machines. It also made me hate Mississippi. I was really hoping GIMP would have some trick up its sleeve to surprise me, but it did not. And this is not far from my past experience on Windows.
Maybe (hopefully) it's better on Linux... one sec...
Mint in VirtualBox: GIMP 1st run: 5 s GIMP 2nd run:2 s.
Hahaha. That just tickles me. It appears GIMP is the iTunes of graphics programs on Windows. Down the rabbit hole...
Within Windows Subsystem for Linux (Ubuntu with vcxsrv as X Windows server) GIMP 1st run (with some bash errors related to OpenGL): 43 s. GIMP 2nd run (still some errors): 5 s.
It's like they anti-optimized the thing for Windows.
Take away: I'll never recommend GIMP to a Windows user again. Linux? Sure, if the UI doesn't bother you, enjoy! But, it's a miserable experience on Windows compared to every other option and should only be used if a particular function can only be performed in GIMP.
Like the way it pops up dialog boxes for most of the geometry manipulation tools, pulling focus from the canvas where you have to hit Enter to commit changes. If I'm dragging with the mouse, I want to simply drag it into position quickly and be done, not search for an OK button or swap focus. It's fine to add some form of fine tuning, but that's usually done with either a menu command or a tool information bar within the GUI. It's functionally the same thing as asking "Are you sure?" at the end of every action.
It violates basic UI assumptions, so unless you use it constantly, you're going to forget which tools operate on which elements in which state. Other graphics programs follow logical flows that are intuitive - select layer, select area of picture, copy, select new layer, paste = selected area in new layer. GIMP wants to buffer every action with an intermediate state that allows some further manipulation, but then provides no intuitive way to take advantage of that state. Like the retention of selected areas as 'floating selections' (or whatever they're called) in the layers list.
Here's a rough analogy: You crack an egg on the edge of a bowl. GIMP wants you to hold onto the egg because you might want to smush it in your fingers or whisper a secret to it now. If you try to put it in the bowl, GIMP ignores that there's a 99% chance that the recipe doesn't call for eggshell bits, so your first attempt results in having to clean out the bowl and crack another egg. Then you realize you first you have to specify that the shell should be removed, with options to change the angle of the shell crack relative to the bowl and the height from which the egg innards are released. Completed, you choose to crack another egg, but because you had not changed focus from the bowl, you proceed to smash the bowl on the counter.
I'm sure there are legitimate cases where having access to those intermediates can be useful, but they shouldn't interrupt every logical step in a common process.
That is probably the most ludicrous statement I've heard today. Being "free software" does not shield you from criticism. And it is not Adobe's fault that the developers of that software do not have a feature. That's entirely on the developers of those programs.
And most people do not care one iota about whether a piece of software is "free" or not. They care if it makes their life easier, and allows them to get a task done. If free software isn't up to it, and a piece of proprietary software is, then the proprietary software is what gets the nod.
"It also suggests you are willing to give up your freedom for a few extra features."
No, it means I want to get my shit done. Just because you are willing to wait does not mean anyone else is.
Can "most people" afford paying $600 yearly fee to edit images? Are "most people" concerned if they fail to make a yearly payment and loose access to all their projects because of vendor lock-in with Adobe's products? Most people I speak to about this are very concerned about this, although they may not know that these problems are caused by Adobe apps being non-free. (But not knowing the cause of a problem doesn't mean the problem doesn't exist)
> Being "free software" does not shield you from criticism.
I agree totally. I just don't think that suggesting proprietary alternatives in this context is particularly helpful as they also lack features of freedom, e.g. sharing, privacy and absence of vendor lock-in.
> And it is not Adobe's fault that the developers of that software do not have a feature. That's entirely on the developers of those programs.
If Adobe freed the source code of Photoshop, how long do you think it would take for missing features to make it into Gimp?
> They care if it makes their life easier, and allows them to get a task done.
... except if the task involves working with other people who may not have access to that software, or may not want to run proprietary software, or may not want to give up their privacy or freedom.
Considering Adobe's Creative Cloud is pretty popular, it seems that yeah, they can. But that's irrelevant to whether a piece of software is FLOSS or not.
"although they may not know that these problems are caused by Adobe apps being non-free"
How dare those bastards ask to be paid for their work!
"I just don't think that suggesting proprietary alternatives in this context is particularly helpful as they also lack features of freedom, e.g. sharing, privacy and absence of vendor lock-in."
In this context, nobody gave a shit about that. They wanted the features that let them get their work done. If FLOSS doesn't have it, it doesn't get used. End of story.
"If Adobe freed the source code of Photoshop, how long do you think it would take for missing features to make it into Gimp?"
Irrelevant, cause that's not going to happen. If the Gimp developers aren't good enough to make those features themselves, without copying off of Adobe, perhaps they should work on something else.
"... except if the task involves working with other people who may not have access to that software, or may not want to run proprietary software, or may not want to give up their privacy or freedom."
So, nobody? Photoshop is the industry standard. If you're not using it, you're the one who's the odd person out, not them.
Totally fine with people charging for their work. Fact is, however, that for half the population of the world, Adobe CS's pricing $50/month is at least 2/3 of their total earnings[1]. By that definition, it's just wrong to say that for "most people" it's an valid alternative to Krita or Gimp.
[1] https://www.dosomething.org/us/facts/11-facts-about-global-p...
If that were to happen, it stands to reason that Gimp would never get any new features, ever again.
RawTherapee and DarkTable are both excellent (I prefer DarkTable), and in some cases have much better plugins and algorithms than Adobe has to offer in Lightroom or Photoshop. Which one works best mainly depends on your preferred UI set-up, but I'd say they're easily at Adobe's level:
[1] https://github.com/GNOME/librsvg
One could just use OpenGL like Blender does to great success, maybe add something like Kivy on top. "Native feel" is not really the most desirable feature for a graphics app, especially since both (1) Linux users are used to non-native looking stuff (even Qt in GTK or the other way around is always at least a bit off) and (2) most native-looking apps, especially the GTK ones, have horrible UI/X, hence users don't tend associate "native looks" with "uuh, this is gonna be a smooth experience" (like Windows users are used to).
(Also, anything that move people even the slightest bit away from Gnome and GTK is a win my book ;) They may feel like the standard now because of Ubuntu, but imho the people working these technologies are totally tone-deaf to what constitutes good UX/I and they are moving further and further in the totally wrong direction... Thank god we have KDE/Qt, and Xfce if you need some "retro but sane" experience in the Gtk camp.)
Also I thought GTK3 was pretty decent cross platform but I may be overly hopeful here.
Stabilizing in what way? I did all the artwork for my last company over the course of ~6 years and can't remember the last time I had a crash. Debian packages.
You'd honestly be surprised by what Aegisub can do. I recently got into fansubbing, mostly as a typesetter, and I'm honestly floored by some of the things I've been able to pull off. I can add text to scenes and make it look like it's part of the video... my favorite effects are the really subtle ones, and it always feels fulfilling whenever I pull something off.
Rotation along the X, Y, and Z axes with the ability to place the origin point anywhere (even off screen!), shear on both the X and Y axes (so useful!), gaussian blur with the ability to blur the border separately from the text (blur is the single most important part about getting text to blend into an image, and it's also a great way to generate glow effects), clipping effects, etc. There's even a drawing mode to work with arbitrary shapes instead of text. And on top of that Aegisub comes with support for scripting effects in Lua, which helps smooth over some of the rougher spots.
I think the next time I want to shoop some text onto an image, especially when I need to match blur and rotation, I might just convert the image to a short video, load it into Aegisub, typeset some text, and screenshot the video.
Disclaimer: I freely admit that I'm a freak and that this probably won't be what most people want. But I find it amusing that a subtitling studio can actually do decent vector graphics, especially with text.
Have fun implementing everything it can do from scratch in your “hipster” language that nobody is going to want to touch two years from now when everyone’s looking for excuses to work in the next hot new language.
I'm disabled and can't work a good paying job to afford the commercial alternatives.
I am also forever a student in learning new things.
Remember Gimp --> GTK? And Gimp is still limping on Gtk2 and blurry on each and every scaled Wayland installation.
Who knows, maybe GIMP will jump straight to Gtk4.
There's a good roundup of happenings in 2017 in GIMP here:
https://www.gimp.org/news/2017/12/31/gimp-and-gegl-in-2017/
[EDIT] There are links in the above doc to sponsor GEGL developers [/EDIT]
It's encouraging that GEGL is seeing use in other projects and receiving patches, maybe the GEGLisation isn't too far from completion.
We are simply not in a position to do centralized crowdfunding.
meanwhile Krita: "We can probably postpone worrying about Wayland on Linux for another year" https://krita.org/en/about/krita-roadmap-2017-2019/ >_<
That being said, I have a Monoprice tablet I use through libinput. I used the wacom driver for a bit but libinput gives me enough to do what I want (I have to say having to do matrix math to constrain the screen size so the tablet is in 1 to 1 correlation to the screen was a bit rough on the UX front). Libinput is pretty feature complete on tablet support in the basic use cases, it is just missing a bunch of the tunables from xsetwacom. The Digimend project has done an amazing job providing driver support for a lot of non-wacom tablets as well. It will be interesting if its Wacom tablets that are the pain point on Wayland. I don't have or know anyone with one to confirm or deny if the libinput support is up to par.
If I had the confidence to know what I'm doing (and I don't, I've never touched the Kwin / Wayland / input stacks on Linux at all) I'd try wiring in tablets to Kwin, but you probably want that first - which is still at least a release or two out at best - before you even think about trying to port your KDE Painting app to Wayland.
Why not provide a --no-tablets-i-just-want-wayland option that would not force the X backend?
So turns out it's easy to turn off X11 in the build files https://github.com/myfreeweb/krita-wayland/commit/8ab43e3e1c... and it does work on Wayland now!
If Krita can replace Photoshop for the most part, it's a good alternative on Linux.
Also, if you can put up with the UI for the plugin, the G'MIC filters contain some cool stuff. You can do frequency separation and content aware fill.
Congrats on 4.0. Great work.