Jump to content

kirk23

Members
  • Posts

    611
  • Joined

  • Last visited

Everything posted by kirk23

  1. I haven't tried this but the video itself don't give me any incentive to even bother trying. It perfectly illustrates my disappointment in AI tools for textures . Totally useless . Even so simple subject as brick texture is visually repeating like hell. Bricks are deformed etc. And all this before the fact that color RGB textures is a last thing you care about when create any material for games or movies. It's depth/height texture that is a base to everything else. A thing you derive color , normal, roughness and everything else . A core thing of any material. And AI so far are pretty bad with generating height . Just try Adobe Sampler AI. I am really looking for de-lighting AI but so far haven't seen anything good. Removing lighting and shadows from photogrammetry source images is what could be a perfect AI purpose but I have yet to see anything useful. Well, actually not from source images , they would instantly turn to be useless for photogrammetry after that , rather from final texture. Any modern albedo/color texture should be deprived of any shadows or shading information at all. It should look very flat and incontrast , just a hue variations and that's all. And all AI tools I have seen so far just do ancient old hipass on them , nothing good. Perhaps it's not even an AI issue, rather people who code it. I have to talk with graphics programmers a lot at my job and usually it's like a talk in between blind and deaf. Programmers who spent their whole life codding visual stuff often don't see an absolutely obvious things . That is always make me kind of stuck since I often even can't find right words to explain something because its so damn obvious and self revealing . Same reason probably why all the non-AI tools for creating CG content are looking like been created by mad scientist . You spend your whole life trying to figure out their crazy logic and deal with their monstrosity . I once heard a story that still amazes me. An artist turned businessmen in US invented lead paint tubes in 19 century . A metal known since antiquity. Before artists used pig bladders to keep their paints from drying . Doing small holes in such bladder evry next day. And they did it for DAMN five if nor more centuries . I very much hope AI wouldn't take five centuries to be useful
  2. So far I used a few AI tools : Simplygon - nope, it's still easier to do lods manually. Simplygon, while now a part of Unreal, still couldn't work efficiently or does it for a few typical subjects only. AI selection in Photoshop - nope , works ok only for kittens and faces, nothing I ever needed. Couldn't recognize a thing beyond those few subjects . Other AI tools in Photoshop is totally useless for me . Ai keywording services for DAM soft. - do a huge mess of keywords . Same issue with recognizing subject as with AI selection in Photoshop. Works fine for people faces and that's all. To bad I never needed that. I have been subscribed to what now Unity art engine for a year. Probably the best AI tool for CG and textures for the moment . Hardly used it at all during the year because other then for a few very specific tasks it was pretty inconvenient. Slow , not good enough preview, always doing anything but what you actually wanted. Lacking any perception how repetitive the final material is looking. I preferred to just use patch tool in Photoshop instead of all that cool AI material regeneration. As much as I hate Substance Designer it still miles ahead and much more useful without any AI even considering its pixel processor and Fx nodes are surely codded by mad scientists. AI material from photo generator in Adobe Sampler. Well , for high quality materials not a chance. Works fine only on very limited subjects and the soft is inconvenient as hell either. Videogames materials have lots of limitation you have to constantly find workarounds for. Not a chance AI could do it for you so far. AI unwrap in Substance painter, Never worked right. Works like a rookie doing its first UV unwrap in Blender . Totally useless. In a word after all that fuss I am so disappointed with AI . My impression it's good only for deep fakes and that's all.
  3. I se I stopped visiting things like that couple decades ago. IMO it's just a visual information noise . I agree it's a kind of fan to create . I instantly recall a nice smell of paint from my old art school, a cool feeling of easy and limitless creative freedom . But now I think about it just that : an information noise. And where AI soon would be dominating probably Now try to do a good looking tree models for a mobile game. From speed tree software for example, a best available. Looks like crap , right . textures looks like crap too. it's where you to need your brain, need to recollect all that boring descriptive geometry courses , vector 3d math and that short 3d graphics theory course in your art school you thought you would never need in your supposed glorious art career etc. IMO it's where true creative and inventive power necessary and where AI wouldn't help you at all. So far at least. Years after my most regret is that I myself have paid not enough attention to all those short ancient geometry courses in art school they considered one of "free arts" . My thoughts was it's not I am going to use ancient Romans mathematics and geometry to build Pantheon dome or something, neither architecture was my specialty. Surprisingly ,years after It turned out it's a knowledge I had a good use of ahead of anything else.
  4. Few years ago I was super exited about AI . Hoped it could help to get rid of lots of annoying routine. Would make human interaction with a computer much more simple and let us focus on creative tasks. Nothing of this have happened since. I admit AI perhaps succeeded in a few very narrow tasks like coloring b&w images and de-noising but mostly AI is a few useless toys , not helping a bit. In many cases it rather annoying like AI selection always selecting something you don't need really . Same with AI usage in other fields of computer generated graphic: AI LODs , AI unwraps , AI 3d modelling etc where it's often stupid like a rock and also rather annoy you than helps. Pretty much like AI support in your bank. My impression it will take hundreds of years till we get any helpful AI really and it dangers only those kinds of jobs where you couldn't stress your own brain at all. You indeed should get rid of such job asap . Including some brainless art generation tasks. Like those where you could call anything an art and that in fact was just a happy and fan random canvas smearing or random photo or fonts compositing . When it comes to something where a good taste, a good sense of nuances , something of a true inventive nature like 3d animation, video games and movie graphics, fashion and costume design, something with clear understandable criteria where you have to find a way to overcome limitations , any field where you should solve unsolvable and hopeless tasks and nobody but you could invent how to do it , in such fields AI is still totally useless and been a huge disappointment during last decade. Any content creation software is still tremendously inconvenient and a huge pain in your ass to use , and at the same time not that flexible at all. It all looks like mad scientist nightmare , totally artist unfriendly , illogical and over-complicated. The adjective "convenient" next to "software" still sounds like dry water , an oxymoron basically . I wish developers would focus on this issue first.
  5. yes, I meant something like this http://www.cgchannel.com/2019/06/tandent-vision-science-launches-lightbrush-cloud/ I had this soft on mac . It never worked that good as on the picture and now is dead I believe. Something AI could do probably
  6. I'd like an AI that could separate albedo color form illumination factor , remove unwanted shadows, increase resolution , recreate Z depth . But still haven't seen any of such things in any other programs that could do it good enough to be of a real help. Even AI selection that could be really helpful works very so so in Photoshop for example and doesn't help a bit really. Rather annoy you like AI technical support in a bank. The whole AI neural networks thing is a total disappointment through recent years. Other than deep fakes it looks like incapable for anything useful. So I wouldn't waste time and focus on gazillion very basic missing things they could improve without AI involved. The soft still has so huge gaps in features I still can't use it for my job and have to use Photoshop to fix files from Affinity photo.
  7. Well. if windows explorer could write keywords not only for jpg and Finder could do hierarchical tags it could be well enough probably without any special DAM . But my guess it's nothing Affinity team could do about.
  8. My guess OP mean something with basic key-wording , rating, metadata and a search based upon , something at Adobe Bridge level at least. Not those horrible file managers both Mac and Windows couldn't make useful for decades for some uncertain reason. But I definitely prefer it would be on OS side rather than Affinity one.
  9. Nvidia has a free standalon soft available to convert files to DDS . Perfectly same as Photoshop plugin. But of course it's an extra pain in your a... by saving a second time each time. And we need first to have that zero alpha black holes multiplied on rgb colors fixed IMO before even think about using AP for textures
  10. Basically a layer rasterizing content of a certain group. To use the rasterized result elsewhere. This would allow to use linked and embedded layers as sources . We could also be able to have complex dynamically calculated masks instead of copies of huge stacks of layers. It would simplifies the main layer stack a lot. Would make it same flexible as any node based editor . To speed up calculations it could be working on demand only as an option , with "update linked " button or something , if too many layers would make it too slow . I'd love also a certain color code for chain image in layers panel/pane. The whole pixel content is red, opasity only is blue etc. Also please , add transforms linking as an option for the linking . It's so much missing there. It's actually the first thing that should have been getting this whole linking idea.
  11. I'd love we have it working same way Pattern view works in Photoshop. Instead of destructive way currently only for a single layer
  12. Layercomps is a great tool in Photoshop. Together with scripts it allows to keep track and save multiple versions of same document automatically. by just single button . While in Affinity it's all same annoying routine. if layercoms are patented or something just make overlapping slices at export persona keeping each its own set of layers . Or just make " select same tagged /named" inverted command so we could select all layers except certain tag /name and switch them off. Ideally in all linked, embedded docs too in sync . Would allow to get rid of so many tedious save /export routine . That export persona doesn't help me a bit currently
  13. We already have live displace filter and it could be a perfect replacement to any manual mesh warp doing basically exactly same thing : warping the image . But weirdly it works all the way wrong vs same thing in evry other soft. Doing kind of opposite shifting on the left and right side. Not sure why. All they need to do is fixing that displace filter and make it sample form 3 channels of "layers behind" Red and green channel to get shift direction /vector and blue channel for a distance of shift. Sure it could be an hour of code tweak since otherwise it's all already working.
  14. The ability to open layered exr files is cool but missing one important thing. We need cryptomatte tool to produce masks , preferably interactively working . Better as procedural filter with sampling capability that would produce masks by clicking on something only when necessary. Would make it another competitive advantage over Photoshop.
  15. The way I could displace/distort pixels in direction set by RG of some vector /normal map and distance from Blue value . In a word a sampling mechanism from another layer /image or rather groups of layers. That only thing would turn it into really procedural image editor . Maybe not as advanced as Substance Designer but I would instantly switch to Affinity anyway. I so much tired of that nodes only things and all the quirks of those pro tools where I really need just simplicity and convenience.
  16. I work as a texture artist and hate all Adobe tools I have to use. Substance designer , Sampler , Painter . Affinity Photo and Designer always seemed a decent alternative whenever you need a quick fix , easy job quickly done, one click macro execution. it's a perfect mix of simplicity and flexibility with it's non destructive linking and layers system. Beyond that of Photoshop. But I miss one thing . An ability to link transforms of the layers. And have to use annoying same tag/name selection that even doesn't have a hot key. it's important since I usually have layers representing height/ normal map/ roughness/ and color values and absolutely need them be always in sync . Whatever groups they might be part of . See Photoshop for example. I tried to use a workaround with symbols containers and pixel content linking but it starts to turn into same insane complexity I hate Painter and Substance Designer for. I want just a click and forget solution . Or maybe you just allow as to add extra channels whenever we need them. Or just make a 16 bit CMYK mode. or rather RGBA+ height/depth. Or just make a checkbox that would select and transform all same named layers in sync whatever group or parent layer they could be related to.
  17. Left side is how I save the file in Photoshop , right side is what it turns to in APhoto . Whenever the alpha is perfect zero it multiplies RGB values to that zero. It makes APhoto unusable for gamedev for example. Just make it keep original values.
  18. In opened tga file for example . If alpha is not perfectly 0 it works just fine but once it's zero the soft multiply RGB values to zero.
  19. 1. 16 and 32 bit CMYk mode. 2. layer compositions 3 wrap around mode ( working not on a single layer but rather for whole document) 4. Gazillion live filters all being live by default . Like height to normal map for example working correctly not in 32 bit images only Through years non stop Krita development made it kind of more advanced image editor than both Photoshop and Aphoto . it's sometimes slow and crushy still I am amazed how advanced and flexible it is. Please borrow some of those features from them.
  20. I'd rather want something like in Photoshop were you can switch documents by special script . like copy layer named "A" from previously selected document and paste it in current document. Photoshop is able to now an order of documents you opened.
  21. Yes,but stamp tools is an ancient concept . It doesn't do anything cool like matching/fusing cloned piece into new surrounding . I have no idea why somebody still use it when there are patch tool, heal tool, content aware move in Psh and inpainting . My guess it still there for historic purpose mostly, an atavism of a past.
  22. It's a huge improvement in Photoshop . Makes in-painted areas so much less repetitive visually. Would be nice to have flip option in patch tool too like Photoshop content aware move tool can . I believe it's nothing really hard to do , basically an extra math operation in the process and for patch tool is just a gizmo option
  23. It's a reason I still pay Photoshop subscription. We could have so many use of extra channel
  24. Just tried this feature in last Rebelle version. Not a big difference really. Rebelle is nice natural media simulator indeed , well ahead of Corel Painter. But the purpose of such simulators always eluded me. For a digital art who cares if it resembles traditional one or not. It's still digital and your skills never translate to actual watercolor or oil/canvas . For sketches and illustrations Photoshop is totally ok and such things like Rebelle are so lacking in basic compositing tools, non-destructive anything it looks like just a toy to please uself sometimes to recollect your art student years decades ago.
×
×
  • Create New...

Important Information

Please note there is currently a delay in replying to some post. See pinned thread in the Questions forum. These are the Terms of Use you will be asked to agree to if you join the forum. | Privacy Policy | Guidelines | We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue.