the plugin doesnt show for me in the list of plugins, i tried the plugin importer (2 times, the second time said it was already installed) and krita is on the C drive, also using Krita 5.2.1, anyone know another fix? Edit: (FIXED )installed the plugin to C:\Program Files\Krita (x64)\share\krita\pykrita then ran krita as Administrator.
sorry i missed this, that's interesting very glade you got it working. im on Windows 10, I'm wondering if you are on Windows 11 and if multiple user accounts on the pc? as in win 10 its C:\Users ame\AppData\Roaming\krita
@@DrFatBearthat's so odd, i checked mine and the directory share is there, whatever cause make it different is beyond me though, but clearly other had had this so ill pin it.
i have installed the plugin as above what is next step because ran as administrator but it does't launch does it mean i cant use it or is there alternative solution to run public server or ?
Great video, I believe you showed us a great work process, I am already running locally Stable Diffusion and created my own models and LORAs to speed up my process, but actually you nail it and I totally agree with you that running all in Krita give us a better UI and allow us to work better and faster, great work, thank you for sharing this to all of us.
Superb video, love the calm voice, excellent walk-through and unequivocal opinion - fantastic. Just discovered Krita yesterday, if I can use this for my workflow then Krita will be getting a decent donation.
Thank you for this video. I just installed it into Krita, and it frankly is a freakin' marvel! I have a 12Gb 3090 Nvidia card, and a 1024x1024 image takes about 10-15 seconds to generate, so it's REALLY fast. The UI is actually richer and has more possibilities than what Photoshop offers for money, and the quality of the generation is miles ahead of Adobe's. Mind blown!
dude, thats pure gold! Just brought adobe photoshop/lightroom for 70 bucks and now there is a free way, actually using stable diffusion for quite a bit, awesome thanks!
@@CINECINEMAS need more details, but i am thinking those with the few day old krita 5.2.2 might be having issues, i put the 5.2.1 on my cloud to download to see if that helps the plugin show up when in the dir
about 8min i mentioned putting plugin into folder and launch krita, at the 9 10 mark i remembered if you have not used krita you need it to run it once for the folder to be created, But yes definitely for anyone that doesn't have the appdate krita folder needs to start krita, close krita, then place the plugin in the folder that should now be there, then re-launch krita. hopefully i did not confuse to many, dyslexic so no scripts when i do the videos, bit rougher than the pro tubers.
@@klaurcschwackerberg1880 there are no credits or limitations, i am not sure what you mean by 300? krita with Ai diffusion is 100% free, no limits on generations, and no credit system. because it runs on your local hardware. I'm not sure why you think or are saying this?
that's cool that they integrated that. the installation could be easier but it works and doesn't use the cloud so it's all good. i installed by running the phyton script installer directly from inside krita
I am using a GTX 1070 and watching this, thinking: This must be in real-time on a powerful machine. And all of a sudden I hear my GPU mentioned! YESSS!
What an amazing tool. I was always lacking some utilities when generating but there's everything i need here, controlnet scribble live generation is awesome and the magic eraser can help fix artifacts or unwanted details, and paired with the use of custom models ? hell yes.
And don't even bother with getting eyes and other details righ on lower res. Start with the general composition, then solve details on higher res. The details might even solve themselves as you upscale and do some img2img passes.
Is there a way to remove and add to an image. Specifically an image of a sideboard in a Studio room that just have the wall and floor.... And then add different elements like vase, painting, plants , sofa, magazines etc.
www.reddit.com/user/Auspicious_Firefly/ This is the plugin creators reddit, I noticed under the inpaint post there where people talking about methods. I myself found in Adobe how use Stable Diffusion to when trying to replace a background something it would not work so I painted the background a solid colour like green then hilghing put park to get it too work. I'm finding that trick works in Krita Diffusion, they mentioned in the reddit the AI Diffuses the selection and uses that information of static colours to create something, they recommend to put put a blob of colour etc. I'll be testing on my backdrop today to see how changing aspects of a image works.
thank you for sharing, but I have some problems, when I use some of my model, it just pop up with "Server execution error.'model.diffusion_model.input_blocks.0.0.weight" is anyone know what happen?
could be a few things, i mentioned in this on a error that might be the same, worth a look, otherwise is it all models and what gpu etc and settings ua-cam.com/video/xh-4YUr1F4U/v-deo.htmlsi=JahpQo0-l0Ehk471
Wow! That was a great and informative demonstration. I was trying to find a way to install Photoshop with the "generative fill" tool, but What I've realized from watching these videos, was that adobe has changed things, and made it harder or impossible for people to download these Ai filters for photoshop to try them for free. It forces them to pay to be able to try their generative fill". Thank God I found your video. But it is a little bit confusing with all the installation steps and models you are referring to. We are not that techie as you are 😉. But I will try. You have a new subscriber to your channel. 👍
Adobe is better on the front of easy. but krita Ai if you have the hardware defiantly has more tools and ability, i love it, I had issues learning the install myself but it was worth working it out.
Does any here, have Automatic 1111 and ComfyUI installed already, and know how to point Krita to an existing directory of models and the ComfyUI location?
he has a how too link a already installed comfyai in there git and reddit, i think they walk someone on a mac os through the step. github.com/Acly/krita-ai-diffusion/blob/main/doc/comfy-requirements.md www.reddit.com/r/StableDiffusion/comments/16kaymk/inpaint_outpaint_in_krita_any_resolution_no/
Handy tip. When going to File Explorer, you can always just use the ENV variable %appdata% to get to this folder, hidden or not. Just type exactly %appdata% into address bar.
There is trouble in Ai image generation section and it is showing error in connecting with server . I have KRITA 5.2.1 version. Can you please help me to resolve it.
For some the models where corrupted during the download or at least some. including myself, I resolved it as per this video link. also I have had a issue where in the setting of the plugin when i hit connect it not connect, for me shutting down my computer or restarting depending on OS setup resolved the issue and then I could get it to connect. my server dir reads C:\Users ame\AppData\Roaming\krita\pykrita\ai_diffusion\.server I have seen some accidental change that when viewing the folder. hopeful one of these help ua-cam.com/video/xh-4YUr1F4U/v-deo.htmlsi=e59vVhvpV5dVgT59
Thank you very much for your reply, may I ask for your email or social media account so that I can show you where I am having problem. @@streamtabulous
I'm gonna throw a question at you, and I'm not sure it applies to Krita in particular. I'm an AI noob who just started a couple of weeks ago. What I would very much like to do, is sort of "overlay" ideas onto existing images. I would like to take a photo of a setting here in my neighborhood, like my street where I live, and then I would like to slap "themes" ONTO that image. Like for instance cyberpunk, futuristic, fantasy that sort of thing. I would then like to sit and tweak these themes and play with the various seeds while KEEPING the perspective intact, so that I can crank up the strength of the image generation quite high without distorting the overall perspective of the photo. I tried generating these themes in Krita, on top of a shot of my living room, and at low strength the perspective is kept, but as I go higher it starts to skew as the process takes greater liberties with the composition. Can you install some sort of controlnet for Krita so that the composition is kept? I guess you could manage that in Comfy or A1111, but what about here?
Yes the control net is when you assign the layer and select image etc that makes the AI use the layer image as the reference information. Also if you check my videos you will see one fully controlling what you want. I'll link in a tick as can't get it while typing. But it basically there a few ways to do it, use the image and have it converted to line art so select line art then type the style you want like futuristic cyberpunk dystopia and it would re render the image to the prompt.
thanks so much for the amazing tutorial! im so glad to have such powerful tool for free! thanks to your video i found out it exists! i already made multiple fun things!! (by the way im a new follower! i love your content!)
Cloud is ok for transferring files like in email and that's it, working offline with software is always better and safer and no company can hold anyone as a hostage.
I prefer local and not on a cloud. Also don't have to worry about there servers going down etc. I find Acly's AI diffusion far better that Adobe's gen fill. And yes I noticed when I used Adobe for the short time you agree to your image being able to be used by them. Don't know if it's still like that but it's a worry as I fixed some family photos and I'd hate to image that they have them, I'm guessing it's for AI training though since now Adobe and even Stable diffusion are no longer training from copy righted or people's art etc. I'd image when you use most cloud AI and even SD has cloud version available that they now have enough generated image there simply training from those.
@@laylasmart yes i know I'm agreeing. i hate cloud stuff, even with reading issues due to dyslexic the agreements seem so sketchy, also not cheap as i mentioned years sub you can buy hardware, same with storage etc.
Awesome video, thanks sir. I used krita but connect to A1111 to improve my workflow. This one in the video will help me a lot. I only have a question, is it possible to run it with only 6B of VRAM?
I have a AMD GPU, and need some help on how to install and run Ai Generator in krita. Also i see a option for running the local server on CPU. So please do a step by step guide, Thanks
Gday fellow Aussie! This looks fantastic! Looks like something I need, however i only have an iPad Pro. Is there a way to download to it? I’m not. I’m not very tech savvy 😢😢😢. Cheers from Central Coast NSW ❤️🇦🇺
Can you help me. I dont have the krita folder in C:\Users\myusername\AppData\Roaming. Even when i have shown to show hidden files. The app is just installed in the Program files folder in C as Krita (x64)
@@petblower1023yes needs to be run once first. Remember needs 27gig of free space for full installation. Can be installed to a different drive I'm told.
@@streamtabulous Yes..i guess it takes a whole day to install all? I got some errors while installing due to that i turned of my antivirus (dont know if that was the reason) But i have tried now again. So far only "Core components" is installed and "Stable diffusion 1.5". Its on "downloading Juggermanut XL" now. Have been sitting with this for some hours now gosh.
@@petblower1023 I actually Think since this video went up there server for the download source might be under strain as seems many are finding it very slow, I think it was 2h for me with my 50Mbps making sure sure nothing else was downloading. But yes some have said 5hours for them. Id say maybe your antivirus monitors the port for incoming data, so data passes through the antivirus, so possibly disconnecting it caused glitch where that layer of security was interrupted, interrupting the data. the sdxl model is 6gig, the sd1.5 2.5 gig ish.. so there take a a bit.
I was thinking about switching to Comfy from using a1111 a little while. Through the process in this video, do I now have proper Comfy installed on my system? How do I run it?
Yer making work flows for Comfy is a headache, have to make heaps of different ones. I like that it's all done for me in the add on for the program and the fact that I have all the paint tools clone tools right there so I don't have to move back and forth from paintshop pro etc. You can link to a installed compfyUi from the plugin that uses your work flow, and the creator says if your on a Mac that's how to get it going as they based it on there own system that's Nvidia. So by that having comfyUi installed might give more user control but they also mentioned there a bunch of add on for compfyUi that need doing also to link it to There Krita addon. I like the simplicity of using their installer and no headaches myself. That said, I do also have CompfyUi Nightly potable to play with that won't conflict with with this. So I'd recommend both. I think the portable would be a go for you too that way you don't break there changes to the work flow. I believe you can mode the bat file in the potable to point to a single model directory. To save space. I have A1111, easy Diffusion, compfyUi that iv yet to work out, and now this Krita with Addon and there installed CompfyUi.
🤔 You might get away with adding a layer change the layer to depth and press the 3 stars. Since that's made to generate a depth map that can be used to help render images but generally it has the look of a ZoeDepth. Or rather a Depth map, and same method but will need to use a sd1.5 model but select Normal that makes a heat map, that's as close to marigold unfortunately but not really good enough. There the only methods that come to mind right now.
although I hate to admit it, what Krita is doing is the future of AI image generation. I love A1111 but it needs to be more intuitive and shouldn't feel like learning a programming language lol
Oh Krita hate AI and are against it, i have not update krita in-case they try and block the add-on, as Krita is open source it means people can make add-ons for it, ACLY did just that and is the sole Guinness of the Ai add-on. but yes adobe and ACLY have it correct this is how the tool is really helpful and beneficial in speeding up productivity
Was a mistake I made, the Diffusion setting I dropped to 2 and 1 from the defaults 7 cause it, worked it out after the video on changing objects. I put a note in the discrimination of that video regarding it. Bugger up on my account played with the settings when I first put it in and never rest them. I'll have to add that note to this video too. I'll try out the in painting model for sure and see how that goes with different settings.
go to sheared in link the download is there, place in to models/clip_vision/SD1.5 github.com/Acly/krita-ai-diffusion/blob/main/doc/comfy-requirements.md
I'm currently playing around with this Krita plugin...funny tool. I can draw, so I generate pictures based on my sketches, sometimes adding text commands. I have an RX6800xt and it works great. If the sketch is quite detailed, it generates an image based on it in a few seconds.
so basically Krita interface is somewhat similar to Photoshop, and is it really free!!!, if it has all the tools of photoshop then i assume it may come with some charges!!! or limited use or pay per use or trial for starters.
Krita its 100% Free there no catch no subs no limitations of any sort. The only this that could be seen as a catch is you need the hardware to run it, while Krita doesn't need much the AI really needs a Nvidia Graphics card in my opinion. in this videos its running on my Nvidia GTX1070. its completely offline once you have downloaded what you need. and all downloads are Free and open source.
Yes in the git they talk about how to do that. From the quick glimpse they mentioned a few add ons you need to do in ComfyAi and how to do it. It's recommended for Linux and Mac users that way also. But I'm not sure how that effects the custom work flow they have, at same time certainly gives you more power too control the comfy to your style.
Awesome! I have a powerful notebook gamer with a RTX3060 with 16mb RAM. First, It takes too long to process with effects and I really don't know why... Second, in my case, the button GENERATE appear EXPAND when I click to expand the image and I don't have the folder SERVER. BTW, I did ALL steps that you teach, and these errors are in my program... Please, can you help me dude? Thank you so much for all!
the layout issue is probably a resolution issue and Krita itself, laptops often have a slightly odd resolution, if you can change the resolution test that. the slow would likely be the canvas size opened is big, the larger the canvas/image the longer rendering takes, try lower first 1000x1000 even 628x628
You should let Mac users know that if they are getting a "xcrun" error this is why! They don't have Xcode installed or the Command Line Tools package is outdated or incompatible.
I don't have mac to test stuff, its very helpful to those that you put this as I know it will help others. will copy your notes and add to the pin list. thank you
for generative fill you need a good graphics card, id recommend all models personally but you can select just the control nets you want. and you can go back later and install ones you left un selected
I have enabled it in Python plugin manager and restarted Krita many times but still "Ai image generation" isn't showing in dockers list 😢, please solve this problem
perhaps try the old Ai diffusion, and older krita. but should be open image, setting, configure krita, python plugin manager {tick Ai image diffusion) apply restart open image, settings, tick show dockers if not ticked, then settings dockers tick Ai image generator.
Sir, A bit of help and advice here? I installed and extracted the application 9:00 then launched Krita again. But, NO sign of the AI Image Diffusion on the Plugin Manager list. Even went back and double checked how and where I extracted the app file. Why would it not be showing up on my list?? It should show up above Assign Profile to Image, but it refuses to appear. Help???
Is it a new install of Krita, I noticed the plugin maker mentioned the version think mine is 5.2.1, that be the first thing I'd check second is the extrated plugin should have two things in the appdata py folder (dyslexia so short handing) make sure the extract plugin is not a single folder if it is go in cut the two extrated out of it and past it in the py folder.
@@streamtabulous My version of Krita is 5.2.2. In the appdata py folder I have 2 items, the ai_diffusion file folder and the ai_diffusion desktop file. Plugin still is not showing up even though, yes, I did restart Krita several times. [EDIT] I have it now!!! There is another way around this issue: Python Plugin Importer. I found this-- ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ Go to Tools ‣ Scripts ‣ Import Python Plugin…, find the *.zip file and press OK. Restart Krita. Go to Configure Krita ‣ Python Plugins Manager, find the plugin and enable it. Restart Krita. Now the plugin should be available. ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ Not exactly like this direction, but even easier. NOW I am ready to PLAY! 😄
@@ijustlovebritcoms omg what a task, will have to copy and paste what you did in the information later for others that have that issue, very interesting on that happening and I wonder why. did you have git and python pre installed, wonder if something conflicted. certainly glade you found a way around it, and glade you posted it as i am sure it will help others
@@streamtabulous I did have an error that led me to the process of installing ComfyUI, a lengthy process I wasn’t expecting but it’s OK….I must go to bed but will investigate in the morning… Maybe that was why the app was hiding, I have Python but didn’t have the UI… [Edited] All seems in good order, my next step will be trying out the new toy!
I only just discovered Krita and would like to use it. But i cannot get the AI plugin to show up after dropping it into the pykrita folder in appdata What version of the ai plugin should be used with Krita 5.2.1? Should I use 5.2.2 I am using SD 1.5 thru ComfyUI but i would like to use Krita as It has better GUI Can you recommend settings or make a video on what is working right now with clean results Thanks for your hard work providing use with informative videos
5.2.2 of krita with 1.17.1 is what i am running sometimes people have had to restart there pc, there a pin list of things others encountered and there work arounds. be sure to open blank image, settings, configuration krita, plug in manager tick ai image diffusion, restart again if needed
Great video! thank you!! one question... I already have SD ComfyUI locally installed... is there a way to make it work with Krita? (withou having to install everything again. seems like waste of hard drive space) Thank you!
yes you can point it to that installed copy. the plugin maker says to install the following as the plugin wont work unless it sees these. github.com/Acly/krita-ai-diffusion/blob/main/doc/comfy-requirements.md
Paid a month sub for Adobe and canceled when I ran out in under 3h using it to help in the free photo restorations I do. Most renders where just white. Has so many errors in Adobe gen fill.
Might be a version conflict. It was plugin 1.9 and krita 5.2.1 At the moment it's plugin version 10.1 and krita 5.2.2 so both are updating quickly. The first thing I'd recommend is checking those versions. I image plugin 10.1 is due to the errors in krita 5.2.2 hard to tell. But since the version changes over the week there been a few with errors. So caused by image size too big, start 512x512, some models are corrupt and getting civitai model and setting that up has fixed the issue. There is a log file under the plugin that might help with some hint where there error might be happening.
Krita/ComfUI cannot handle if you ask it for a plain background or to remove the background after you've selected the subject. I don't see how it's ever going to compete with Adobe.
it does do that. some models are bad but 60% of the one i use do that easily when doing portraits they want to look like professional photographer on plane or gradient background
@@streamtabulous I don't believe that. I spent hours the last week and Krita/ComfUI couldn't handle it. Not even negative prompts could stop it forcing faces and lamps into the background.
@@ericneo2 what are you doing and what models etc, should not need negative prompt, with a bad model you can invert subject selection flood adjust slider for strength to 70% the render a perfect gradient background
@@streamtabulous Standard fresh install of Krita, ComfUI and SD1.5. Selection and invert only changed if the generated content was inside or outside the selection. It would outright ignore negative prompts.
@@ericneo2 sd1.5 is likely the issue, i moved to sdxl for better results across the board, but some models do force backgrounds, but there are professional photography models that work great on gradient backgrounds, Photoshop is better at it yes but there model is massive so with stable diffusion you unfortunately need to find the right models to use. but if you select subject invert selection, new layer flood fill the layer then reduce strength of the Ai generation so it pays attention to the flooded background lower strength less changes to the flood fill, but this method works for feathering in and i find i can get effects like blue sponge painted wall, gradients colour etc. its more playing around to adobe but constantly getting better and free if you already have hardware to run it. i think negative prompts is broken in krita atm seen some reports to acly some is model training as use to be wd14 now there all new language for prompts that via some stops negative prompt but i think thats more sdxl. but hopefully the above helps. i should do some short tip videos when i can
Why does mine takes a long time to generate it's images, i have a RTX 4070 and it's taking forever, even after increasing it's power in the settings :c
might be something to do with overall system 🤔 mine amd3600 with 32gig 3600mhz ram asus 550 board, also have cuda tool kit installed. these are my krita ai settings ua-cam.com/video/QE9-5uzUhrE/v-deo.htmlsi=Z-LVf7siB4lMgo4o
Is there any way to install everything on the D: drive instead of the C: drive? I can't really afford to add over 20GB of data onto my C: drive right now. From looking at where default folders are stored (at 6:38 in the video), it seems like it may be a requirement to install it on the C: drive.
Yes,, I believe if you change the path: Connection server path it will install to the location you put. I tested and moved the directory of mine and pasted in the a difference drive, pointed the server path to that drive and location of the folder and it worked. Note: it did run much slower in the mechanical drive compared to the SSD.
Hi , may you can help: The ai generater was working fine yesterday. Today it says: Server execusion Error: Insufficent funds ... asking for Tokens So its not free? Or i did something wrong?
uninstall and reinstall, it your running it locally on your own hardware and gpu its 100% free. the beta cloud gpu is only rent option and im not aware of how that would warn. but local is 100% free.
Well all seemed to install well, but so far generating doesn't finish, the progress bar moves all the way to the right (or so it looks) but never pops in the image viewer. Is there a way to see the comfyui server? I've run comfyui separately before and seeing the messages as it's busy can be helpful. I'll try again tomorrow
I have not linked it to my comfy yet, There is no way to see the server, agree would be a handy option, under the connection there is a view log though. I could not find the bat in the custom dir to launch the version the plugin uses, believe they may have done some stuff to stop people breaking it. I have had that issue of it seeming to stop, I resolved it by reducing my image size and found too big it would freeze. maybe open a new image at 512x512 and test
@@streamtabulous Thanks, agree that i think i had my image size to render too big,opened a new image at 512x512 and it rendered. and yes, looking to the server log file was helpful too to get those msg's
@streamtabulous thanks sadly all I have is a 1650 :( haha just working with what I've got. Right now things are tight so not able to upgrade yet. It is interesting tho to me that prior to this experiment with kritas plug-in I explored using comfyui and it wasn't as slow as it seems to be with the krita setup. Wasn't fast mind you, but trying to use the live drawing is nigh unusable for me. Possibly has to do with the particular nodes, models it uses. I'll keep tinkering tho as the potential is really cool, especially for an open-source tool, props ro the krita team!
@@zoemorn I know what you mean, I imaging its the double handling plus the built in workflow, I would be curious of linking to a installed comfyAI and seeing if it makes a difference. also do you have the cude drivers installed? its recommended in most forums i have been to on us running slower stuff, I have cuda in and it made a massive difference to a1111 as per my video on that, but i feel the cuda driver link is important all round for this stuff, and be sure to set krita to 6gig. krita is open source im impressed by it, the AI addon is one person
first canvas size, large canvas uses lots more ram, set performance ram to the highest ram use that does no go above the amount of ram your gpu has, and also what gpu, with the gtx1070 i needed to do the mod in this video ua-cam.com/video/FzvaNV9fQ3o/v-deo.htmlsi=VRLHnDPQuok_xZ3p
little lost on what you mean? krita installs whats needed, do you mean one is not downloaded and has a error, or do you mean installing a upscaler thats not in the krita Ai install package. you can PM on Instagram with photos, I can look normal see if normal etc, if you youtube (Acly) has a golden icon he is the creator, I leave comment on his latest videos and he has responded or fixed stuff based on that, pretty awesome with how busy they must be. But give me more details version number plus above etc,
i have tryed Stable Diffusion as local server, but its much too slow on GTX1070 with Ryzen 2700x on win 10. You have to wait for hours to create a picture. So it's not suitable for everyday use for me. What kind of hardware did you need to make this run correctly? Did it need a RTX GPU or did you need a SSD drive for usable performence?
AMD 3600 , 2gig 5200rpm mechanical drive with gtx1070 8gig, have to remember the image canvas size maters a lot, start with a image canvas of 680x680, second is model and steps, etc cfg 2 steps 12 to 15 lower steps faster it will work, to low some model wont work and make a mess, so find the balance, made video today to show and help you and others for my settings etc
Feels like me with Australia internet. 26 or 27 gig is the whole thing. I put it on a thumb drive the put on my other pc the one I record the video on. And I'll be keeping it backed up just incase.
@@N269 what errors? i put up a link to 5.2.1 as krita updated to 5.2.2 that might have compatibility issues, waiting on feedback. the app data might still be in the folder. unfortunately my cloud is not big enough to put the full server download or id definitely do that
install cuda tool kit, always recommend that for nvida uses just hit the cross on the page dont need a log in, And image canvas size if doing a full image smaller is faster so start with a 512x512 canvas and slowly increase size till you find a canvas sive that works best in speed for you. and Model selection, I use SD1.5 models on my old system rather that SDXL, as SD1.5 are faster for live paint.
some trouble shooting from others listed. will add as people solve stuff they run into. Krita needs to be run once for app data folder to be created. @zoemorn 9:10 you must restart Krita for it to show up in Dockers. thank you @ijustlovebritcoms Plugin still is not showing up even though, yes, I did restart Krita several times. [EDIT] I have it now!!! There is another way around this issue: Python Plugin Importer. I found this-- ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ Go to Tools ‣ Scripts ‣ Import Python Plugin…, find the *.zip file and press OK. Restart Krita. Go to Configure Krita ‣ Python Plugins Manager, find the plugin and enable it. Restart Krita. Now the plugin should be available. ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ Those with COMFY installed the plugin maker said yes you can link it to your Comfy, Has how on there git and reddit, i think they walk someone on a mac os in the reddit, I believe you need to do a few steps to do so. github.com/Acly/krita-ai-diffusion/blob/main/doc/comfy-requirements.md www.reddit.com/r/StableDiffusion/comments/16kaymk/inpaint_outpaint_in_krita_any_resolution_no/ Mac users: from Decster, You should let Mac users know that if they are getting a "xcrun" error this is why! They don't have Xcode installed or the Command Line Tools package is outdated or incompatible.
@@streamtabulous I've been playing around with it quite a bit since last night. Give it a week and I'll upload some stuff. If animate_diff gets implemented into the krita animation workspace, that's a game changer. If you could get the stable diff prompt travel and key frames to output to krita key frames, the bar to homemade animations just became so low you can walk over it... figuratively speaking.
Agree I don't use drift but it be nice to see add ons like the live paint that reads the colour, I trying to add that myself to there krita instead comfyAI but the server would not start after doing it oddly, removing the server started. Hopefully more work is don't on like and support for krita v5.2.2 as it might not be compatible by some feedback so I put my 5.2.1 on my cloud. Im guessing linking it it won't use custom stuff as the plugin needs the icons and python code to point to use those features. Unfortunately my brain couldn't learn code. Guess the dyslexia speech spelling stuff might affect that learning
I have a acer aspire 7 with Nvidia GeForce gtx 1650 (4gb) + intel Uhd (4gb) … 16gb ram.. i5 12core, base speed 1.70GHz... Can i use it for at list 500 * 500 pix image project …? I am a game dev, and I want to use this as a starting ground of my graphics/background... please give an answer... where I live internet is kind of pricey … I don't want to waist 15gb for nothing.... please give an answer ?
in the comments there is a list of peoples issues and how they fixed it, try this one Plugin still is not showing up even though, yes, I did restart Krita several times. [EDIT] I have it now!!! There is another way around this issue: Python Plugin Importer. I found this-- ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ Go to Tools ‣ Scripts ‣ Import Python Plugin…, find the *.zip file and press OK. Restart Krita. Go to Configure Krita ‣ Python Plugins Manager, find the plugin and enable it. Restart Krita. Now the plugin should be available.
@@streamtabulous When i do this and go to configure Krita and manage python i see the ai diffusion option but i cannot enable ot or disable it because it is checked on and un clickable and sim. the text itself is dim as well. it says "NO MANUAL" at the bottom when i do click on it. :(
Don't know if anyone done videos on that you would need to search it. The makers reddit has a person that put it on a mac. Need linking to a installed comfy and not installing the auto DL one
I'm running on a4 amd cpu with rtx1070 8gig, i did a long rant video on hardware and speeds on a intel all in one HP pc. gaming systems are the basic rull of thumb and nvida as mentioned in video is the best,
@@streamtabulous You mean running AI server on a laptop with no graphic card... Still, I wander if it works with the Intel Iris Xe graphic processor on laptop or it works only with desktop with GPU for this heavy lifting.
@@laylasmart not sure contracts on cloud gpu rentals or how they work. i mean anything image related. stable diffusion will run on a potato, not good but can, i talked about someones system in this rambling video speeds he got what his system is and speeds after doing so setting. ua-cam.com/video/mRz-c4eC0jg/v-deo.htmlsi=Ij69EDks4yZewDpq
@@laylasmart his was intel i9 integrated graphics but 8gig for that integrated graphics but need low steps and cfg. and around 2min wait times to 4min, but stock setting where a hour at 30 steps
@@haki1894 the plugin makers reddit is in pinned link, they say AMD GPU supported via DirectML on Windows, but i don't see if special drivers are needed or what amd gpus
Unfortunately not, he could possibly do it later, but if he did it would need to run on a rented GPU via cloud. Also I'm not sure how much coding difference there is so that could be a complication to make it work. Right now it's mac and pc only
@@sudipadhikary9648 probably try something like AI playground Web UI 100 a day free, use Firefox app in desktop mode to it thinks its loading the PC WebUi as ut only works in desktop Web UI mode.
www.interstice.cloud/plugin the link show how to set it up, its free running and using your local hardware. i imagine you have it set up wrong and maybe have it set to the newly added cloud gpu rental that i have no idea on.
@@streamtabulous I compared the video to a human audible frequency sweep. The tone I'm hearing seems to be somewhere in the 10kHz - 20kHz range, maybe close to 18 kHz. This video plays it at about 0:34 - ua-cam.com/video/PAsMlDptjx8/v-deo.htmlsi=MBL9ml_l02_BuAGr&t=34
According to audio analysis software of your video there is a very notable spike in the spectrogram at about 15.1 kHz, which matches with the sound playing in the previous video.
@@sourdface no idea there is nothing on and noise cancellation both mic plus obs, maybe UA-cam transcoding issue but i can't here it and opening the og there is no wave frequency showing in the audio so no idea.
yer found out much later its c+++ but thank to Python tool called SIP, which makes it possible for python speak in C++ and talk to Krita and thank goodness.
I don't know mac's, there are reddit forums on krita AI that would be more helpful than myself. i do know from the first video on krita i did mac users ran it with good results. from Acly this is the install and whats need On Linux/Mac: Python + venv must be installed recommended version: 3.11 or 3.10 Apple M1/M2 supported via MPS on macOS MacOS: Usually ~/Library/Application Support/krita/pykrita this information is here, I hope it works for you. github.com/Acly/krita-ai-diffusion did aslo say to start at 512x512 with sd1.5 models at 10 to 12 steps and cfg 2 if it runs well slowly try other image canvas sizes 628x628 / 628x800 etc
OK I tried the new toy but nothing happened. Got this error message: AssertionError Python 3.10.7: C:\Program Files\Krita (x64)\bin\krita.exe Tue Dec 12 11:28:32 2023 A problem occurred in a Python script. Here is the sequence of function calls leading up to the error, in the order they occurred. C:\Users\xxxxx_\AppData\Roaming\krita\pykrita\ai_diffusion\ui\generation.py in apply_selected_result(self=) 274 275 def apply_selected_result(self): 276 self.model.apply_result() 277 278 def apply_result(self, item: QListWidgetItem): self = self.model = self.model.apply_result = C:\Users\xxxx_\AppData\Roaming\krita\pykrita\ai_diffusion\model.py in apply_result(self=) 298 299 def apply_result(self): 300 assert self._layer and self.can_apply_result 301 self._layer.setLocked(False) 302 self._layer.setName(self._layer.name().replace("[Preview]", "[Generated]")) self = self._layer = None self.can_apply_result = False AssertionError: __cause__ = None __class__ = __context__ = None __delattr__ = __dict__ = {} __dir__ = __doc__ = 'Assertion failed.' __eq__ = __format__ = __ge__ = __getattribute__ = __gt__ = __hash__ = __init__ = __init_subclass__ = __le__ = __lt__ = __ne__ = __new__ = __reduce__ = __reduce_ex__ = __repr__ = __setattr__ = __setstate__ = __sizeof__ = __str__ = __subclasshook__ = __suppress_context__ = False __traceback__ = args = () with_traceback = The above is a description of an error in a Python program. Here is the original traceback: Traceback (most recent call last): File "C:\Users\xxxx_\AppData\Roaming\krita\pykrita\ai_diffusion\ui\generation.py", line 276, in apply_selected_result self.model.apply_result() File "C:\Users\xxxx_\AppData\Roaming\krita\pykrita\ai_diffusion\model.py", line 300, in apply_result assert self._layer and self.can_apply_result AssertionError
Download a civitai model and try that, there errors lines at the model ny my old eyes, I had a error if you watch the adding a civitai video that was fixed putting models in, i wonder if there where corruptions downloading models, worth rulling out quick and easy for a model, sd1.5 are normally 2gig ish
@@streamtabulousI wonder if my 3 yr old computer GTX 1060 is just too slow. Also maybe installing all the ComfyUI models was a bit much for it. But I will try the Civitai model and let you know how it goes 🙂
@@ijustlovebritcoms i do know someone that was running it on a gtx1050 they said it was slow but worked. someone asked if it could be installed on a different drive so I tested and moved my server data that was installed in a old mechanical drive, that made the renders take much longer, makes my wonder if the model thats part of he way AI thinks is no cached to memory but on the SSD its much faster. that said ye old gtx1070 is good fast bit memory. but save the adobe sub money and use it to upgrade is my current advice to everyone including myself
LOL I am often a victim of bad timing, wish I had known about what Krita can do before subbing to Adobe, oof… Well, getting the SD app running with any success might take a while. I need to ask you where and how to download the CivitAI model files? HuggingFace is a bit confusing to me. And the CivitAI website refers to JSON files, etc., I am not sure how to proceed.
@@ijustlovebritcoms to poor for adobe sub especially when they bumped the prices up so high. Adobe implementation of stable diffusion is fantastic though, anyone with cash to blow id say go for it lol CivitAi.com links in the video on how to add them, just download the checkpoints, thats the models, the json file is the krita metadata link thats created when you add a new model profile. I'll be sure to add a video on how to use civitiai website and download as its probably handy for some.
There plugin page is what linked me to the plugin of AI Diffusion so that ironic and they have links to more AI tools than this one. In fact the spot removal tool that was implemented to remove people and objects is a basic AI implementation in krita so that's very odd statement when they have AI in it. There plans are to keep it basic for low end uses. But they certainly don't discourage AI.
Also it's ignorance that AI steals art it doesn't. I talk about this lots on how AI works. There is no images in the models. And at current Stable Diffusion use NO copyright images to train the AI, training is not loading images into models, the AI looks at the image and uses Diffusion and mostly language to understand what its looking at and scattered image fragment. Adobe did train on copyright art namely famous ones Van Gogh etc. Some Krita groups are likely just staying out of the AI discussion as they are open source and the artist and extremely non intellectual in what AI is and how AI works and in such are angry over it because they done understand it, and the last thing Krita won't to do is upset those artists because they rely on there donations. blog.adobe.com/en/publish/2020/02/27/copyrights-in-the-era-of-ai
depends no hardware and size your doing, its 6sec to 10sec on the 8gig rtx3060 and 15sec to 20sec sdxl on the gtx1070 8gig ua-cam.com/video/QE9-5uzUhrE/v-deo.htmlsi=65ciqDXxbF2di9lC a lot faster than my old adobe gen fill.
@@Babiii-g8n sadly yes it would be slow, needs gaming GPUs nvidia is best, gtx670, gtx960 gtx1060, and newer rtx is super fast. there is a beta cloud rental option that you have to pay though but much cheaper that adobe by far, there information on that on aclys page if thats of interest. also full install of krita ai diffusion is 26gig.
@@streamtabulous i minimized my losses and just uninstalled it lol installed maya and 3dsmax instead lets stop supporting AI gonna take away your jobs one day
@@Babiii-g8n disagree there many professionals using it reducing pressure the meat deadlines, as long as laws are in place to protect jobs there is no issues only benefits. being disabled it helped me massively. plus as in my recent video if you want it to do a style or consistent character you need to train it requires the commission of a artist to build that data set to do so. Ai can't replace a artist its limited to its training. end of the day its only a tool and its the hands of the person using it that is responsible for the results.
set memory to 6g or lower, drop step to 10 to 12, guidance to 2, and lower image size to 628x628 ish. and set gpu to MLC , stable diffusion is built on NVIDIA cuda so sadly not as good on other cards, but there are AMD users reporting faster speeds than my old GTX1070
What? I mentioned i'm running it on a old gtx1070, there is a guy running it on a dell with intel onboard gpu i talk about in another video on setting it up for old low hardware. I mention the year cost of abode you can get a RTX GPU and it be one off that way rather than years of payments, also on Acly plugin page he has the setup for cheap cloud gpu renting that it works with also. No kidney needed to be sold, just lower your size to 628x628 or 512x512 use sd1.5 lower steps and guidance and it will run on a potato
@@abdelrahmanahmed1766 yer need 4gig vram Min. 1gig would be a issue unfortunately. If you don't mind the second hand market and it's a case pc you could look at second hand gtx970, Gtx1070 both do well and are cheap. Can get whole second hand pc with those cards cheap well cheaper that a years subscription to Adobe. Otherwise on Acly plug it has a guide to cloud gpu and how to do that. Apparently it's cheap like $2 a hour and you only pay for what you use. It's how most people make the models by renting cloud gpu. I have not tried it, you could have a fiddle and see how you go. Being the plugin and krita are free it's worth fiddling with. That said! for the less computer minded I recommend AI playground that's free for 100 renders a day and if you put live preview on that's free, so you can save the final renders for ones your really like. And screen shot the lesser ones. To get more. AI playground is good.
@@streamtabulous tnx for replying I will stick now to ALPACA cloud AI processing for now unfortunately second hand gtx1070 in my country worth a salary of two months 🫠
I think Generative fill in Krita is not 100% free , because in accounts you can see you have only 300 credits to use , it does not say if that is per month or per day , or per year. or anything about it.
there are no credits or limitations, i am not sure what you mean by 300? krita with Ai diffusion is 100% free, no limits on generations, and no credit system. because it runs on your local hardware
read through pinned post, ensue 25gig free space when installing and start with 512x512 image size. but read pinned posts its from users who had issues and there fixes.
there is, its like saying dont bother sculpting because 3d printers can do it, there always something special about hands on. I still do some hands on dispute the pain and limitations of my disability's. also you need artists to create images to train a model to do what you want, I trained a model on my art, I mentioned i won't copy it made something new but cool and unique
the plugin doesnt show for me in the list of plugins, i tried the plugin importer (2 times, the second time said it was already installed) and krita is on the C drive, also using Krita 5.2.1, anyone know another fix?
Edit: (FIXED )installed the plugin to C:\Program Files\Krita (x64)\share\krita\pykrita then ran krita as Administrator.
SAME PROBLEM
sorry i missed this, that's interesting very glade you got it working.
im on Windows 10, I'm wondering if you are on Windows 11 and if multiple user accounts on the pc?
as in win 10 its C:\Users
ame\AppData\Roaming\krita
I am on Windows 10 with a single account
@@DrFatBearthat's so odd, i checked mine and the directory share is there, whatever cause make it different is beyond me though, but clearly other had had this so ill pin it.
i have installed the plugin as above what is next step because ran as administrator but it does't launch does it mean i cant use it or is there alternative solution to run public server or ?
Great video, I believe you showed us a great work process, I am already running locally Stable Diffusion and created my own models and LORAs to speed up my process, but actually you nail it and I totally agree with you that running all in Krita give us a better UI and allow us to work better and faster, great work, thank you for sharing this to all of us.
Superb video, love the calm voice, excellent walk-through and unequivocal opinion - fantastic. Just discovered Krita yesterday, if I can use this for my workflow then Krita will be getting a decent donation.
Thank you for this video. I just installed it into Krita, and it frankly is a freakin' marvel! I have a 12Gb 3090 Nvidia card, and a 1024x1024 image takes about 10-15 seconds to generate, so it's REALLY fast. The UI is actually richer and has more possibilities than what Photoshop offers for money, and the quality of the generation is miles ahead of Adobe's.
Mind blown!
dude, thats pure gold! Just brought adobe photoshop/lightroom for 70 bucks and now there is a free way, actually using stable diffusion for quite a bit, awesome thanks!
plus you won't get the against policy warnings and unlimited generation, save the monthly adobe sub to put towards a new gpu later
HI, can you please help me to download i am having problem in adding the defusion with krita
@@CINECINEMAS need more details, but i am thinking those with the few day old krita 5.2.2 might be having issues, i put the 5.2.1 on my cloud to download to see if that helps the plugin show up when in the dir
9:10 you must restart Krita for it to show up in Dockers. thanks for the video though! i'm your 300th like. :)
about 8min i mentioned putting plugin into folder and launch krita, at the 9 10 mark i remembered if you have not used krita you need it to run it once for the folder to be created,
But yes definitely for anyone that doesn't have the appdate krita folder needs to start krita, close krita, then place the plugin in the folder that should now be there, then re-launch krita.
hopefully i did not confuse to many, dyslexic so no scripts when i do the videos, bit rougher than the pro tubers.
ill definitely add your note to the pinned comment for sure to hopeful stop any confusion, thanks
@@streamtabulous what about the 300 credits that are mentioned in the account info ? it doesn't sound like free free but limited.
@@klaurcschwackerberg1880 there are no credits or limitations, i am not sure what you mean by 300?
krita with Ai diffusion is 100% free, no limits on generations, and no credit system.
because it runs on your local hardware.
I'm not sure why you think or are saying this?
@@klaurcschwackerberg1880 what account info? there is no account at all.
that's cool that they integrated that. the installation could be easier but it works and doesn't use the cloud so it's all good.
i installed by running the phyton script installer directly from inside krita
I am using a GTX 1070 and watching this, thinking: This must be in real-time on a powerful machine. And all of a sudden I hear my GPU mentioned! YESSS!
yep, I really need to do a video titled gtx1070 you can do it ..
What an amazing tool. I was always lacking some utilities when generating but there's everything i need here, controlnet scribble live generation is awesome and the magic eraser can help fix artifacts or unwanted details, and paired with the use of custom models ? hell yes.
hey thank you for this! I wanted to ask how do we add different models for landscape models / car models etc..
sir, it is very helpful! i'm not that often write a comment but it is really mind-blowing so thanks a lot! and from now on i subscribe your channel !
So nice of you
For generation on a weaker graphics card you should start with 512x512, or 512x768. Then upscale and inpaint the details.
Great tip in general, even for stronger cards. Explore the possibilities with low res then upscale and fine tune.
And don't even bother with getting eyes and other details righ on lower res. Start with the general composition, then solve details on higher res. The details might even solve themselves as you upscale and do some img2img passes.
Is there a way to remove and add to an image. Specifically an image of a sideboard in a Studio room that just have the wall and floor.... And then add different elements like vase, painting, plants , sofa, magazines etc.
www.reddit.com/user/Auspicious_Firefly/
This is the plugin creators reddit, I noticed under the inpaint post there where people talking about methods.
I myself found in Adobe how use Stable Diffusion to when trying to replace a background something it would not work so I painted the background a solid colour like green then hilghing put park to get it too work. I'm finding that trick works in Krita Diffusion, they mentioned in the reddit the AI Diffuses the selection and uses that information of static colours to create something, they recommend to put put a blob of colour etc.
I'll be testing on my backdrop today to see how changing aspects of a image works.
Does it work without any "login" process and that too completely locally?
yes it runs completely off your Computer, there no logins
thank you for sharing, but I have some problems, when I use some of my model, it just pop up with "Server execution error.'model.diffusion_model.input_blocks.0.0.weight"
is anyone know what happen?
could be a few things, i mentioned in this on a error that might be the same, worth a look, otherwise is it all models and what gpu etc and settings
ua-cam.com/video/xh-4YUr1F4U/v-deo.htmlsi=JahpQo0-l0Ehk471
13:10 what does "adding your own models" mean?
mine is No Manual Available, please help to fix it
This is what I was hoping would happen. Great information.
Wow! That was a great and informative demonstration. I was trying to find a way to install Photoshop with the "generative fill" tool, but What I've realized from watching these videos, was that adobe has changed things, and made it harder or impossible for people to download these Ai filters for photoshop to try them for free. It forces them to pay to be able to try their generative fill". Thank God I found your video. But it is a little bit confusing with all the installation steps and models you are referring to. We are not that techie as you are 😉. But I will try. You have a new subscriber to your channel. 👍
Adobe is better on the front of easy. but krita Ai if you have the hardware defiantly has more tools and ability, i love it, I had issues learning the install myself but it was worth working it out.
WOW!, there's a big learning curve but I'll work on it. Thanks for the tips mate 👍👍.
Does any here, have Automatic 1111 and ComfyUI installed already, and know how to point Krita to an existing directory of models and the ComfyUI location?
he has a how too link a already installed comfyai in there git and reddit, i think they walk someone on a mac os through the step.
github.com/Acly/krita-ai-diffusion/blob/main/doc/comfy-requirements.md
www.reddit.com/r/StableDiffusion/comments/16kaymk/inpaint_outpaint_in_krita_any_resolution_no/
Handy tip. When going to File Explorer, you can always just use the ENV variable %appdata% to get to this folder, hidden or not. Just type exactly
%appdata%
into address bar.
There is trouble in Ai image generation section and it is showing error in connecting with server . I have KRITA 5.2.1 version. Can you please help me to resolve it.
For some the models where corrupted during the download or at least some.
including myself, I resolved it as per this video link.
also I have had a issue where in the setting of the plugin when i hit connect it not connect, for me shutting down my computer or restarting depending on OS setup resolved the issue and then I could get it to connect.
my server dir reads
C:\Users
ame\AppData\Roaming\krita\pykrita\ai_diffusion\.server
I have seen some accidental change that when viewing the folder.
hopeful one of these help
ua-cam.com/video/xh-4YUr1F4U/v-deo.htmlsi=e59vVhvpV5dVgT59
Thank you very much for your reply, may I ask for your email or social media account so that I can show you where I am having problem. @@streamtabulous
@@CINECINEMASuse the link to my instagram
I'm gonna throw a question at you, and I'm not sure it applies to Krita in particular.
I'm an AI noob who just started a couple of weeks ago.
What I would very much like to do, is sort of "overlay" ideas onto existing images. I would like to take a photo of a setting here in my neighborhood, like my street where I live, and then I would like to slap "themes" ONTO that image. Like for instance cyberpunk, futuristic, fantasy that sort of thing. I would then like to sit and tweak these themes and play with the various seeds while KEEPING the perspective intact, so that I can crank up the strength of the image generation quite high without distorting the overall perspective of the photo. I tried generating these themes in Krita, on top of a shot of my living room, and at low strength the perspective is kept, but as I go higher it starts to skew as the process takes greater liberties with the composition.
Can you install some sort of controlnet for Krita so that the composition is kept?
I guess you could manage that in Comfy or A1111, but what about here?
Yes the control net is when you assign the layer and select image etc that makes the AI use the layer image as the reference information.
Also if you check my videos you will see one fully controlling what you want. I'll link in a tick as can't get it while typing. But it basically there a few ways to do it, use the image and have it converted to line art so select line art then type the style you want like futuristic cyberpunk dystopia and it would re render the image to the prompt.
ua-cam.com/video/V8ITwgkRXzA/v-deo.htmlsi=C2Fm6wTH3LYkktyH
thanks so much for the amazing tutorial! im so glad to have such powerful tool for free! thanks to your video i found out it exists! i already made multiple fun things!! (by the way im a new follower! i love your content!)
thank you, that's great to read
Cloud is ok for transferring files like in email and that's it, working offline with software is always better and safer and no company can hold anyone as a hostage.
I prefer local and not on a cloud. Also don't have to worry about there servers going down etc. I find Acly's AI diffusion far better that Adobe's gen fill.
And yes I noticed when I used Adobe for the short time you agree to your image being able to be used by them. Don't know if it's still like that but it's a worry as I fixed some family photos and I'd hate to image that they have them, I'm guessing it's for AI training though since now Adobe and even Stable diffusion are no longer training from copy righted or people's art etc. I'd image when you use most cloud AI and even SD has cloud version available that they now have enough generated image there simply training from those.
@@streamtabulous
Exactly!!! That's what offline means. You and the computer and that's it.
@@laylasmart yes i know I'm agreeing. i hate cloud stuff, even with reading issues due to dyslexic the agreements seem so sketchy, also not cheap as i mentioned years sub you can buy hardware, same with storage etc.
Awesome video, thanks sir. I used krita but connect to A1111 to improve my workflow.
This one in the video will help me a lot. I only have a question, is it possible to run it with only 6B of VRAM?
Yes definitely would run on 6gig vram.
I have a AMD GPU, and need some help on how to install and run Ai Generator in krita. Also i see a option for running the local server on CPU. So please do a step by step guide, Thanks
Thanks :) I decided to pull models I had in comfy. I am enjoying this much more than using a1111 for inpainting. (I'm subbing!!)
It's my favorite so far, Krita alone is amazing and with the add-on fantastic
Gday fellow Aussie! This looks fantastic! Looks like something I need, however i only have an iPad Pro. Is there a way to download to it? I’m not. I’m not very tech savvy 😢😢😢. Cheers from Central Coast NSW ❤️🇦🇺
Krita is available on IOS but unfortunately the Plugin is not available for IOS or Android, and is only available on PC at this time.
Amazing content, subscribed!!
Welcome aboard!
Can you help me. I dont have the krita folder in C:\Users\myusername\AppData\Roaming. Even when i have shown to show hidden files. The app is just installed in the Program files folder in C as Krita (x64)
Oh i see that i have to run the krita first before heading over to that folder.
@@petblower1023yes needs to be run once first.
Remember needs 27gig of free space for full installation.
Can be installed to a different drive I'm told.
@@streamtabulous Yes..i guess it takes a whole day to install all? I got some errors while installing due to that i turned of my antivirus (dont know if that was the reason) But i have tried now again. So far only "Core components" is installed and "Stable diffusion 1.5". Its on "downloading Juggermanut XL" now. Have been sitting with this for some hours now gosh.
@@petblower1023 I actually Think since this video went up there server for the download source might be under strain as seems many are finding it very slow, I think it was 2h for me with my 50Mbps making sure sure nothing else was downloading.
But yes some have said 5hours for them.
Id say maybe your antivirus monitors the port for incoming data, so data passes through the antivirus, so possibly disconnecting it caused glitch where that layer of security was interrupted, interrupting the data.
the sdxl model is 6gig, the sd1.5 2.5 gig ish.. so there take a a bit.
I was thinking about switching to Comfy from using a1111 a little while. Through the process in this video, do I now have proper Comfy installed on my system? How do I run it?
Yer making work flows for Comfy is a headache, have to make heaps of different ones. I like that it's all done for me in the add on for the program and the fact that I have all the paint tools clone tools right there so I don't have to move back and forth from paintshop pro etc.
You can link to a installed compfyUi from the plugin that uses your work flow, and the creator says if your on a Mac that's how to get it going as they based it on there own system that's Nvidia.
So by that having comfyUi installed might give more user control but they also mentioned there a bunch of add on for compfyUi that need doing also to link it to There Krita addon.
I like the simplicity of using their installer and no headaches myself.
That said,
I do also have CompfyUi Nightly potable to play with that won't conflict with with this.
So I'd recommend both. I think the portable would be a go for you too that way you don't break there changes to the work flow.
I believe you can mode the bat file in the potable to point to a single model directory. To save space.
I have A1111, easy Diffusion, compfyUi that iv yet to work out, and now this Krita with Addon and there installed CompfyUi.
How would one add say ZoeDepth or marigold depth to Krita? Anyone have any clues there?
🤔 You might get away with adding a layer change the layer to depth and press the 3 stars.
Since that's made to generate a depth map that can be used to help render images but generally it has the look of a ZoeDepth. Or rather a Depth map, and same method but will need to use a sd1.5 model but select Normal that makes a heat map, that's as close to marigold unfortunately but not really good enough.
There the only methods that come to mind right now.
although I hate to admit it, what Krita is doing is the future of AI image generation. I love A1111 but it needs to be more intuitive and shouldn't feel like learning a programming language lol
Oh Krita hate AI and are against it, i have not update krita in-case they try and block the add-on, as Krita is open source it means people can make add-ons for it, ACLY did just that and is the sole Guinness of the Ai add-on.
but yes adobe and ACLY have it correct this is how the tool is really helpful and beneficial in speeding up productivity
i think if use inpainting models the lines problem in out painting will go off
Was a mistake I made, the Diffusion setting I dropped to 2 and 1 from the defaults 7 cause it, worked it out after the video on changing objects. I put a note in the discrimination of that video regarding it.
Bugger up on my account played with the settings when I first put it in and never rest them.
I'll have to add that note to this video too.
I'll try out the in painting model for sure and see how that goes with different settings.
I got it to connect to my comfyui but i'm getting an error about clipvision pytorch_model.bin needs to be in a folder. where do i find that model?
go to sheared in link the download is there, place in to models/clip_vision/SD1.5
github.com/Acly/krita-ai-diffusion/blob/main/doc/comfy-requirements.md
Krita = Crayon in swedish. :) "Kreeta" is closer to the correct pronunciation than "Kritta", but don't sweat it.
Thanks for the heads up!
I'm horribly dyslexic and have reading written and speech issues, civitai i stumble all over saying it too.
eeeeee lol 4th video I'll get it, if my brain does not derp out on me. thanks
awesome content, thanks, helped a lot in my projects
have more stuff coming soon on different way to use and control, just lots of medical stuff in the way
After dropping plugin..open krita then close then open it again it works for me.
I'm currently playing around with this Krita plugin...funny tool. I can draw, so I generate pictures based on my sketches, sometimes adding text commands. I have an RX6800xt and it works great. If the sketch is quite detailed, it generates an image based on it in a few seconds.
is that 512x512, fantastic to read that AMD gpu is working great with it
so basically Krita interface is somewhat similar to Photoshop, and is it really free!!!, if it has all the tools of photoshop then i assume it may come with some charges!!! or limited use or pay per use or trial for starters.
Krita its 100% Free there no catch no subs no limitations of any sort.
The only this that could be seen as a catch is you need the hardware to run it, while Krita doesn't need much the AI really needs a Nvidia Graphics card in my opinion.
in this videos its running on my Nvidia GTX1070.
its completely offline once you have downloaded what you need. and all downloads are Free and open source.
If you already have comfyui installed. Can you connect to it so you don't have two versions on your pc?
Yes in the git they talk about how to do that. From the quick glimpse they mentioned a few add ons you need to do in ComfyAi and how to do it.
It's recommended for Linux and Mac users that way also.
But I'm not sure how that effects the custom work flow they have, at same time certainly gives you more power too control the comfy to your style.
Awesome! I have a powerful notebook gamer with a RTX3060 with 16mb RAM. First, It takes too long to process with effects and I really don't know why... Second, in my case, the button GENERATE appear EXPAND when I click to expand the image and I don't have the folder SERVER. BTW, I did ALL steps that you teach, and these errors are in my program... Please, can you help me dude? Thank you so much for all!
the layout issue is probably a resolution issue and Krita itself, laptops often have a slightly odd resolution, if you can change the resolution test that.
the slow would likely be the canvas size opened is big, the larger the canvas/image the longer rendering takes, try lower first 1000x1000 even 628x628
@@streamtabulous thank for your help, I really appreciate it!
You should let Mac users know that if they are getting a "xcrun" error this is why! They don't have Xcode installed or the Command Line Tools package is outdated or incompatible.
I don't have mac to test stuff, its very helpful to those that you put this as I know it will help others. will copy your notes and add to the pin list.
thank you
Very informative. I want to use it for generative fill, do I need to install all those models data?
for generative fill you need a good graphics card, id recommend all models personally but you can select just the control nets you want. and you can go back later and install ones you left un selected
I have enabled it in Python plugin manager and restarted Krita many times but still "Ai image generation" isn't showing in dockers list 😢, please solve this problem
perhaps try the old Ai diffusion, and older krita.
but should be open image, setting, configure krita, python plugin manager {tick Ai image diffusion) apply restart open image, settings, tick show dockers if not ticked, then settings dockers tick Ai image generator.
@@streamtabulous thanks, I will try this
Sir, A bit of help and advice here? I installed and extracted the application 9:00 then launched Krita again. But, NO sign of the AI Image Diffusion on the Plugin Manager list. Even went back and double checked how and where I extracted the app file. Why would it not be showing up on my list?? It should show up above Assign Profile to Image, but it refuses to appear. Help???
Is it a new install of Krita, I noticed the plugin maker mentioned the version think mine is 5.2.1, that be the first thing I'd check second is the extrated plugin should have two things in the appdata py folder (dyslexia so short handing) make sure the extract plugin is not a single folder if it is go in cut the two extrated out of it and past it in the py folder.
you need to restart krita
@@streamtabulous My version of Krita is 5.2.2. In the appdata py folder I have 2 items, the ai_diffusion file folder and the ai_diffusion desktop file. Plugin still is not showing up even though, yes, I did restart Krita several times.
[EDIT] I have it now!!! There is another way around this issue: Python Plugin Importer. I found this--
~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
Go to Tools ‣ Scripts ‣ Import Python Plugin…, find the *.zip file and press OK. Restart Krita.
Go to Configure Krita ‣ Python Plugins Manager, find the plugin and enable it. Restart Krita.
Now the plugin should be available.
~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
Not exactly like this direction, but even easier. NOW I am ready to PLAY! 😄
@@ijustlovebritcoms omg what a task, will have to copy and paste what you did in the information later for others that have that issue, very interesting on that happening and I wonder why. did you have git and python pre installed, wonder if something conflicted. certainly glade you found a way around it, and glade you posted it as i am sure it will help others
@@streamtabulous I did have an error that led me to the process of installing ComfyUI, a lengthy process I wasn’t expecting but it’s OK….I must go to bed but will investigate in the morning… Maybe that was why the app was hiding, I have Python but didn’t have the UI… [Edited] All seems in good order, my next step will be trying out the new toy!
I only just discovered Krita and would like to use it. But i cannot get the AI plugin to
show up after dropping it into the pykrita folder in appdata
What version of the ai plugin should be used with Krita 5.2.1? Should I use 5.2.2
I am using SD 1.5 thru ComfyUI but i would like to use Krita as It has better GUI
Can you recommend settings or make a video on what is working right now with clean results
Thanks for your hard work providing use with informative videos
5.2.2 of krita with 1.17.1 is what i am running sometimes people have had to restart there pc, there a pin list of things others encountered and there work arounds.
be sure to open blank image, settings, configuration krita, plug in manager tick ai image diffusion, restart again if needed
while downloading the server it shows a error "invalid location: directory is not empty, but no previous installation was found" plz help
needs 25g to install might not be enough space, you can point it elsewhere to install.
the generative fill is very slow.... my PC is very powerful, how do i make it faster?
GPU is whats needed, but using lighting model, lower steps and lower cfg
say 10 steps 3 cfg for example
Great video! thank you!! one question... I already have SD ComfyUI locally installed... is there a way to make it work with Krita? (withou having to install everything again. seems like waste of hard drive space) Thank you!
yes you can point it to that installed copy. the plugin maker says to install the following as the plugin wont work unless it sees these.
github.com/Acly/krita-ai-diffusion/blob/main/doc/comfy-requirements.md
So so useful as usual !!! Can you tell me if i can use KRITA with an AMD graphic card ?
git plugin says yes using mlc, what gen GPUs no idea. being free I'd definitely recommend trying it
Imagine that Adobe recently announced they will be charging credits for every image generation in their firefly AI
Paid a month sub for Adobe and canceled when I ran out in under 3h using it to help in the free photo restorations I do. Most renders where just white. Has so many errors in Adobe gen fill.
My laptop is not NVIDIA, uses Intel IRIS, must use DirectML (GPU), is there a tutorial?
live generate did not show the thumbnail of the generation art
and if i make apply showing to me script error
any solution ???!!
Might be a version conflict. It was plugin 1.9 and krita 5.2.1
At the moment it's plugin version 10.1 and krita 5.2.2 so both are updating quickly.
The first thing I'd recommend is checking those versions. I image plugin 10.1 is due to the errors in krita 5.2.2 hard to tell. But since the version changes over the week there been a few with errors. So caused by image size too big, start 512x512, some models are corrupt and getting civitai model and setting that up has fixed the issue.
There is a log file under the plugin that might help with some hint where there error might be happening.
@@streamtabulous thanks ❤
i will try and tell you
its work
i updated krita
then use 512x512 and its working
@@a7medetman150 that's great to read 🤗
Krita/ComfUI cannot handle if you ask it for a plain background or to remove the background after you've selected the subject. I don't see how it's ever going to compete with Adobe.
it does do that. some models are bad but 60% of the one i use do that easily when doing portraits they want to look like professional photographer on plane or gradient background
@@streamtabulous I don't believe that. I spent hours the last week and Krita/ComfUI couldn't handle it. Not even negative prompts could stop it forcing faces and lamps into the background.
@@ericneo2 what are you doing and what models etc, should not need negative prompt, with a bad model you can invert subject selection flood adjust slider for strength to 70% the render a perfect gradient background
@@streamtabulous Standard fresh install of Krita, ComfUI and SD1.5. Selection and invert only changed if the generated content was inside or outside the selection. It would outright ignore negative prompts.
@@ericneo2 sd1.5 is likely the issue, i moved to sdxl for better results across the board, but some models do force backgrounds, but there are professional photography models that work great on gradient backgrounds, Photoshop is better at it yes but there model is massive so with stable diffusion you unfortunately need to find the right models to use. but if you select subject invert selection, new layer flood fill the layer then reduce strength of the Ai generation so it pays attention to the flooded background lower strength less changes to the flood fill, but this method works for feathering in and i find i can get effects like blue sponge painted wall, gradients colour etc. its more playing around to adobe but constantly getting better and free if you already have hardware to run it.
i think negative prompts is broken in krita atm seen some reports to acly some is model training as use to be wd14 now there all new language for prompts that via some stops negative prompt but i think thats more sdxl. but hopefully the above helps. i should do some short tip videos when i can
Why does mine takes a long time to generate it's images, i have a RTX 4070 and it's taking forever, even after increasing it's power in the settings :c
might be something to do with overall system 🤔 mine amd3600 with 32gig 3600mhz ram asus 550 board,
also have cuda tool kit installed.
these are my krita ai settings
ua-cam.com/video/QE9-5uzUhrE/v-deo.htmlsi=Z-LVf7siB4lMgo4o
Is there any way to install everything on the D: drive instead of the C: drive? I can't really afford to add over 20GB of data onto my C: drive right now. From looking at where default folders are stored (at 6:38 in the video), it seems like it may be a requirement to install it on the C: drive.
Yes,, I believe if you change the path: Connection server path it will install to the location you put.
I tested and moved the directory of mine and pasted in the a difference drive, pointed the server path to that drive and location of the folder and it worked.
Note: it did run much slower in the mechanical drive compared to the SSD.
@@streamtabulous That's good to know. Thanks!
Hi , may you can help: The ai generater was working fine yesterday.
Today it says: Server execusion Error: Insufficent funds ... asking for Tokens
So its not free? Or i did something wrong?
uninstall and reinstall, it your running it locally on your own hardware and gpu its 100% free. the beta cloud gpu is only rent option and im not aware of how that would warn.
but local is 100% free.
Well all seemed to install well, but so far generating doesn't finish, the progress bar moves all the way to the right (or so it looks) but never pops in the image viewer. Is there a way to see the comfyui server? I've run comfyui separately before and seeing the messages as it's busy can be helpful. I'll try again tomorrow
I have not linked it to my comfy yet,
There is no way to see the server, agree would be a handy option, under the connection there is a view log though.
I could not find the bat in the custom dir to launch the version the plugin uses, believe they may have done some stuff to stop people breaking it.
I have had that issue of it seeming to stop, I resolved it by reducing my image size and found too big it would freeze. maybe open a new image at 512x512 and test
@@streamtabulous Thanks, agree that i think i had my image size to render too big,opened a new image at 512x512 and it rendered. and yes, looking to the server log file was helpful too to get those msg's
@@zoemorn work you size up from there, the gtx1070 is good with 624x1000
@streamtabulous thanks sadly all I have is a 1650 :( haha just working with what I've got. Right now things are tight so not able to upgrade yet. It is interesting tho to me that prior to this experiment with kritas plug-in I explored using comfyui and it wasn't as slow as it seems to be with the krita setup. Wasn't fast mind you, but trying to use the live drawing is nigh unusable for me. Possibly has to do with the particular nodes, models it uses. I'll keep tinkering tho as the potential is really cool, especially for an open-source tool, props ro the krita team!
@@zoemorn I know what you mean, I imaging its the double handling plus the built in workflow, I would be curious of linking to a installed comfyAI and seeing if it makes a difference. also do you have the cude drivers installed? its recommended in most forums i have been to on us running slower stuff, I have cuda in and it made a massive difference to a1111 as per my video on that, but i feel the cuda driver link is important all round for this stuff, and be sure to set krita to 6gig.
krita is open source im impressed by it, the AI addon is one person
I've used Krita ai for a while, error message appear recently: Server Execution error: Invalide buffer size: 10.15 GB, what's the solution to solve?
first canvas size, large canvas uses lots more ram, set performance ram to the highest ram use that does no go above the amount of ram your gpu has,
and also what gpu, with the gtx1070 i needed to do the mod in this video
ua-cam.com/video/FzvaNV9fQ3o/v-deo.htmlsi=VRLHnDPQuok_xZ3p
bro I cant install ultimate sd upscale in comfy ui manager. What do I do??
little lost on what you mean?
krita installs whats needed, do you mean one is not downloaded and has a error, or do you mean installing a upscaler thats not in the krita Ai install package.
you can PM on Instagram with photos, I can look normal see if normal etc, if you youtube (Acly) has a golden icon he is the creator, I leave comment on his latest videos and he has responded or fixed stuff based on that, pretty awesome with how busy they must be.
But give me more details version number plus above etc,
i have tryed Stable Diffusion as local server, but its much too slow on GTX1070 with Ryzen 2700x on win 10. You have to wait for hours to create a picture. So it's not suitable for everyday use for me. What kind of hardware did you need to make this run correctly? Did it need a RTX GPU or did you need a SSD drive for usable performence?
AMD 3600 , 2gig 5200rpm mechanical drive with gtx1070 8gig, have to remember the image canvas size maters a lot, start with a image canvas of 680x680, second is model and steps, etc cfg 2 steps 12 to 15 lower steps faster it will work, to low some model wont work and make a mess, so find the balance,
made video today to show and help you and others for my settings etc
ua-cam.com/video/QE9-5uzUhrE/v-deo.htmlsi=ZK8EBYFwXEJhc4M1
Well, I'm going to try this!!!!!!!!! Even if it take 3 days to download!!!!! Been 5hrs and still going.....
Feels like me with Australia internet. 26 or 27 gig is the whole thing.
I put it on a thumb drive the put on my other pc the one I record the video on. And I'll be keeping it backed up just incase.
@@streamtabulousWell, first attemp, errors, errors..... Krita now removed. Thanks!
@@N269 what errors? i put up a link to 5.2.1 as krita updated to 5.2.2 that might have compatibility issues, waiting on feedback. the app data might still be in the folder. unfortunately my cloud is not big enough to put the full server download or id definitely do that
@@streamtabulous
Whenever "Apply" is pressed:-
AssertionError
Python 3.10.7: C:\Program Files\Krita (x64)\bin\krita.exe
Wed Dec 13 08:55:33 2023
A problem occurred in a Python script. Here is the sequence of
function calls leading up to the error, in the order they occurred.
C:\Users\User\AppData\Roaming\krita\pykrita\ai_diffusion\ui\generation.py in apply_selected_result(self=)
276
277 def apply_selected_result(self):
278 self.model.apply_result()
279
280 def apply_result(self, item: QListWidgetItem):
self =
self.model =
self.model.apply_result =
C:\Users\User\AppData\Roaming\krita\pykrita\ai_diffusion\model.py in apply_result(self=)
300
301 def apply_result(self):
302 assert self._layer and self.can_apply_result
303 self._layer.setLocked(False)
304 self._layer.setName(self._layer.name().replace("[Preview]", "[Generated]"))
self =
self._layer = None
self.can_apply_result = False
AssertionError:
__cause__ = None
__class__ =
__context__ = None
__delattr__ =
__dict__ = {}
__dir__ =
__doc__ = 'Assertion failed.'
__eq__ =
__format__ =
__ge__ =
__getattribute__ =
__gt__ =
__hash__ =
__init__ =
__init_subclass__ =
__le__ =
__lt__ =
__ne__ =
__new__ =
__reduce__ =
__reduce_ex__ =
__repr__ =
__setattr__ =
__setstate__ =
__sizeof__ =
__str__ =
__subclasshook__ =
__suppress_context__ = False
__traceback__ =
args = ()
with_traceback =
The above is a description of an error in a Python program. Here is
the original traceback:
Traceback (most recent call last):
File "C:\Users\User\AppData\Roaming\krita\pykrita\ai_diffusion\ui\generation.py", line 278, in apply_selected_result
self.model.apply_result()
File "C:\Users\User\AppData\Roaming\krita\pykrita\ai_diffusion\model.py", line 302, in apply_result
assert self._layer and self.can_apply_result
AssertionError
generative fill is 1024x 1024 res what is resolution with KRITA?
SDXL is 1024x1024, adobe use stability AI so same Resolution.
however with krita and on civitAI there are a few models trained on 2k and 4k.
i need help i have a 3060 12vram but it take slow the generation i mean its not realtime like yours
install cuda tool kit, always recommend that for nvida uses just hit the cross on the page dont need a log in,
And image canvas size if doing a full image smaller is faster so start with a 512x512 canvas and slowly increase size till you find a canvas sive that works best in speed for you.
and Model selection, I use SD1.5 models on my old system rather that SDXL, as SD1.5 are faster for live paint.
some trouble shooting from others listed. will add as people solve stuff they run into.
Krita needs to be run once for app data folder to be created.
@zoemorn
9:10 you must restart Krita for it to show up in Dockers. thank you
@ijustlovebritcoms
Plugin still is not showing up even though, yes, I did restart Krita several times.
[EDIT] I have it now!!! There is another way around this issue: Python Plugin Importer. I found this--
~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
Go to Tools ‣ Scripts ‣ Import Python Plugin…, find the *.zip file and press OK. Restart Krita.
Go to Configure Krita ‣ Python Plugins Manager, find the plugin and enable it. Restart Krita.
Now the plugin should be available.
~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
Those with COMFY installed the plugin maker said yes you can link it to your Comfy,
Has how on there git and reddit, i think they walk someone on a mac os in the reddit, I believe you need to do a few steps to do so.
github.com/Acly/krita-ai-diffusion/blob/main/doc/comfy-requirements.md
www.reddit.com/r/StableDiffusion/comments/16kaymk/inpaint_outpaint_in_krita_any_resolution_no/
Mac users: from Decster,
You should let Mac users know that if they are getting a "xcrun" error this is why! They don't have Xcode installed or the Command Line Tools package is outdated or incompatible.
Thanks, I did end up getting it going. Thank you for the vid.
@@denniskwasnycia635 great to read, probably better on you comfy as you can update etc
@@streamtabulous I've been playing around with it quite a bit since last night. Give it a week and I'll upload some stuff. If animate_diff gets implemented into the krita animation workspace, that's a game changer. If you could get the stable diff prompt travel and key frames to output to krita key frames, the bar to homemade animations just became so low you can walk over it... figuratively speaking.
Agree I don't use drift but it be nice to see add ons like the live paint that reads the colour, I trying to add that myself to there krita instead comfyAI but the server would not start after doing it oddly, removing the server started. Hopefully more work is don't on like and support for krita v5.2.2 as it might not be compatible by some feedback so I put my 5.2.1 on my cloud.
Im guessing linking it it won't use custom stuff as the plugin needs the icons and python code to point to use those features. Unfortunately my brain couldn't learn code. Guess the dyslexia speech spelling stuff might affect that learning
I have a acer aspire 7 with Nvidia GeForce gtx 1650 (4gb) + intel Uhd (4gb) … 16gb ram.. i5 12core, base speed 1.70GHz...
Can i use it for at list 500 * 500 pix image project …? I am a game dev, and I want to use this as a starting ground of my graphics/background... please give an answer... where I live internet is kind of pricey … I don't want to waist 15gb for nothing.... please give an answer ?
Thanks, Very helpfull!!!
why its not working on my laptop its not appear on dockers can you help to solve it
in the comments there is a list of peoples issues and how they fixed it,
try this one
Plugin still is not showing up even though, yes, I did restart Krita several times.
[EDIT] I have it now!!! There is another way around this issue: Python Plugin Importer. I found this--
~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
Go to Tools ‣ Scripts ‣ Import Python Plugin…, find the *.zip file and press OK. Restart Krita.
Go to Configure Krita ‣ Python Plugins Manager, find the plugin and enable it. Restart Krita.
Now the plugin should be available.
@@streamtabulous When i do this and go to configure Krita and manage python i see the ai diffusion option but i cannot enable ot or disable it because it is checked on and un clickable and sim. the text itself is dim as well. it says "NO MANUAL" at the bottom when i do click on it. :(
@@CoreyMcKinneyJr www.interstice.cloud/plugin
Is it mandatory to install all modules or I can instal modules which I prefer?
you can install just what you want, and if need be tick and install others later
Does anyone have verified steps for installing Ai-Diffusion on Krita 5.2.2 on a M2 macbook?
Don't know if anyone done videos on that you would need to search it. The makers reddit has a person that put it on a mac. Need linking to a installed comfy and not installing the auto DL one
cheers. cannot find a video yet but will keep on searching.@@streamtabulous
and hardware requirements for this workflow?
I'm running on a4 amd cpu with rtx1070 8gig, i did a long rant video on hardware and speeds on a intel all in one HP pc. gaming systems are the basic rull of thumb and nvida as mentioned in video is the best,
This is Super Amazing.
26GB on CPU only is too much. Can you make her without cloths or it is limited as the website?
26GB is not that much, PC games are much larger, when you think that your running essentially your own AI sever.
It unlimited does nsfw
@@streamtabulous
You mean running AI server on a laptop with no graphic card...
Still, I wander if it works with the Intel Iris Xe graphic processor on laptop or it works only with desktop with GPU for this heavy lifting.
@@laylasmart not sure contracts on cloud gpu rentals or how they work. i mean anything image related.
stable diffusion will run on a potato, not good but can, i talked about someones system in this rambling video speeds he got what his system is and speeds after doing so setting.
ua-cam.com/video/mRz-c4eC0jg/v-deo.htmlsi=Ij69EDks4yZewDpq
@@streamtabulous
Not contracts of online GPU. Intel iris xe is an imbedded graphic card in their CPU for laptops.
So I wonder how it will work...
@@laylasmart his was intel i9 integrated graphics but 8gig for that integrated graphics but need low steps and cfg. and around 2min wait times to 4min, but stock setting where a hour at 30 steps
Can I make it work with my GTX 1050ti?
i know a guy that was using a gtx1050 6g non ti, so yes, he did say was slow but work.
@@streamtabulousThat's good to know. However, my version is the 4GB one. I might be out of luck.
@@mrlightwriter id try it, stable diffusion the core is 4gig min. it will just be longer. reduce image size to 512x512 and don't use xl models
@@streamtabulous Thank you again for your reply. It does indeed work with my 1050ti 4GB, albeit slowly.
bro I installed everything but it gives me a red error, do you know how to fix this?
I have an AMD gpu, it works when i click use CPU
@@haki1894 the plugin makers reddit is in pinned link, they say AMD GPU supported via DirectML on Windows, but i don't see if special drivers are needed or what amd gpus
Is ai available for android tab?
Unfortunately not, he could possibly do it later, but if he did it would need to run on a rented GPU via cloud.
Also I'm not sure how much coding difference there is so that could be a complication to make it work.
Right now it's mac and pc only
@@streamtabulous could you please suggest me is there any good generative AI available for android?
@@sudipadhikary9648 probably try something like AI playground Web UI 100 a day free, use Firefox app in desktop mode to it thinks its loading the PC WebUi as ut only works in desktop Web UI mode.
Thank you for this.
How do you add tokens please
www.interstice.cloud/plugin
the link show how to set it up, its free running and using your local hardware. i imagine you have it set up wrong and maybe have it set to the newly added cloud gpu rental that i have no idea on.
Great Video! Very useful. Thank you a lot!
very welcome
Is it just me or is there a horrible high pitch ringing in the background?
noise cancelling mic and software, I'm not hearing it, also i dont have a air-conditioner or fan etc so not sure what your hearing 🤨
@@streamtabulous I compared the video to a human audible frequency sweep. The tone I'm hearing seems to be somewhere in the 10kHz - 20kHz range, maybe close to 18 kHz. This video plays it at about 0:34 - ua-cam.com/video/PAsMlDptjx8/v-deo.htmlsi=MBL9ml_l02_BuAGr&t=34
According to audio analysis software of your video there is a very notable spike in the spectrogram at about 15.1 kHz, which matches with the sound playing in the previous video.
@@sourdface no idea there is nothing on and noise cancellation both mic plus obs, maybe UA-cam transcoding issue but i can't here it and opening the og there is no wave frequency showing in the audio so no idea.
@@sourdfacehas to be your system, checked 10 so far including this and nothing. id say its a decoding issue with your system
WOW! Thanks!
Krita is not written in Python
yer found out much later its c+++ but thank to Python tool called SIP, which makes it possible for python speak in C++ and talk to Krita and thank goodness.
Thanks
On Mac what do I need to run AI?
I don't know mac's, there are reddit forums on krita AI that would be more helpful than myself. i do know from the first video on krita i did mac users ran it with good results.
from Acly this is the install and whats need
On Linux/Mac: Python + venv must be installed
recommended version: 3.11 or 3.10
Apple M1/M2 supported via MPS on macOS
MacOS: Usually ~/Library/Application Support/krita/pykrita
this information is here, I hope it works for you.
github.com/Acly/krita-ai-diffusion
did aslo say to start at 512x512 with sd1.5 models at 10 to 12 steps and cfg 2
if it runs well slowly try other image canvas sizes 628x628 / 628x800 etc
excelent video!
It's not free after sometimes ask for token
then your not installing the real Acly version and your not set up local as its fully free and run off your hardware
OK I tried the new toy but nothing happened. Got this error message:
AssertionError
Python 3.10.7: C:\Program Files\Krita (x64)\bin\krita.exe
Tue Dec 12 11:28:32 2023
A problem occurred in a Python script. Here is the sequence of
function calls leading up to the error, in the order they occurred.
C:\Users\xxxxx_\AppData\Roaming\krita\pykrita\ai_diffusion\ui\generation.py in apply_selected_result(self=)
274
275 def apply_selected_result(self):
276 self.model.apply_result()
277
278 def apply_result(self, item: QListWidgetItem):
self =
self.model =
self.model.apply_result =
C:\Users\xxxx_\AppData\Roaming\krita\pykrita\ai_diffusion\model.py in apply_result(self=)
298
299 def apply_result(self):
300 assert self._layer and self.can_apply_result
301 self._layer.setLocked(False)
302 self._layer.setName(self._layer.name().replace("[Preview]", "[Generated]"))
self =
self._layer = None
self.can_apply_result = False
AssertionError:
__cause__ = None
__class__ =
__context__ = None
__delattr__ =
__dict__ = {}
__dir__ =
__doc__ = 'Assertion failed.'
__eq__ =
__format__ =
__ge__ =
__getattribute__ =
__gt__ =
__hash__ =
__init__ =
__init_subclass__ =
__le__ =
__lt__ =
__ne__ =
__new__ =
__reduce__ =
__reduce_ex__ =
__repr__ =
__setattr__ =
__setstate__ =
__sizeof__ =
__str__ =
__subclasshook__ =
__suppress_context__ = False
__traceback__ =
args = ()
with_traceback =
The above is a description of an error in a Python program. Here is
the original traceback:
Traceback (most recent call last):
File "C:\Users\xxxx_\AppData\Roaming\krita\pykrita\ai_diffusion\ui\generation.py", line 276, in apply_selected_result
self.model.apply_result()
File "C:\Users\xxxx_\AppData\Roaming\krita\pykrita\ai_diffusion\model.py", line 300, in apply_result
assert self._layer and self.can_apply_result
AssertionError
Download a civitai model and try that, there errors lines at the model ny my old eyes, I had a error if you watch the adding a civitai video that was fixed putting models in, i wonder if there where corruptions downloading models, worth rulling out quick and easy for a model, sd1.5 are normally 2gig ish
@@streamtabulousI wonder if my 3 yr old computer GTX 1060 is just too slow. Also maybe installing all the ComfyUI models was a bit much for it. But I will try the Civitai model and let you know how it goes 🙂
@@ijustlovebritcoms i do know someone that was running it on a gtx1050 they said it was slow but worked.
someone asked if it could be installed on a different drive so I tested and moved my server data that was installed in a old mechanical drive, that made the renders take much longer, makes my wonder if the model thats part of he way AI thinks is no cached to memory but on the SSD its much faster. that said ye old gtx1070 is good fast bit memory. but save the adobe sub money and use it to upgrade is my current advice to everyone including myself
LOL I am often a victim of bad timing, wish I had known about what Krita can do before subbing to Adobe, oof… Well, getting the SD app running with any success might take a while. I need to ask you where and how to download the CivitAI model files? HuggingFace is a bit confusing to me. And the CivitAI website refers to JSON files, etc., I am not sure how to proceed.
@@ijustlovebritcoms to poor for adobe sub especially when they bumped the prices up so high. Adobe implementation of stable diffusion is fantastic though, anyone with cash to blow id say go for it lol
CivitAi.com links in the video on how to add them, just download the checkpoints, thats the models, the json file is the krita metadata link thats created when you add a new model profile. I'll be sure to add a video on how to use civitiai website and download as its probably handy for some.
Nice
krita team and artists and studios and companies who is support krita not allow to use it for AI
There plugin page is what linked me to the plugin of AI Diffusion so that ironic and they have links to more AI tools than this one.
In fact the spot removal tool that was implemented to remove people and objects is a basic AI implementation in krita so that's very odd statement when they have AI in it.
There plans are to keep it basic for low end uses. But they certainly don't discourage AI.
Also it's ignorance that AI steals art it doesn't. I talk about this lots on how AI works. There is no images in the models. And at current Stable Diffusion use NO copyright images to train the AI, training is not loading images into models, the AI looks at the image and uses Diffusion and mostly language to understand what its looking at and scattered image fragment.
Adobe did train on copyright art namely famous ones Van Gogh etc.
Some Krita groups are likely just staying out of the AI discussion as they are open source and the artist and extremely non intellectual in what AI is and how AI works and in such are angry over it because they done understand it, and the last thing Krita won't to do is upset those artists because they rely on there donations.
blog.adobe.com/en/publish/2020/02/27/copyrights-in-the-era-of-ai
at ua-cam.com/video/FiKR_TrMioI/v-deo.html Your mouse goes behind you. It took me a couple time to really get exactly what you clicked.
Yer fixed in the newer videos
i hope my gtx 1660 ti can
Rtx1660 ti definitely can.
isnt it so slow?
depends no hardware and size your doing,
its 6sec to 10sec on the 8gig rtx3060 and 15sec to 20sec sdxl on the gtx1070 8gig
ua-cam.com/video/QE9-5uzUhrE/v-deo.htmlsi=65ciqDXxbF2di9lC
a lot faster than my old adobe gen fill.
@@streamtabulous im using nvidia mx250 20gb ram
@@Babiii-g8n sadly yes it would be slow, needs gaming GPUs nvidia is best, gtx670, gtx960 gtx1060, and newer rtx is super fast.
there is a beta cloud rental option that you have to pay though but much cheaper that adobe by far, there information on that on aclys page if thats of interest.
also full install of krita ai diffusion is 26gig.
@@streamtabulous i minimized my losses and just uninstalled it lol installed maya and 3dsmax instead lets stop supporting AI gonna take away your jobs one day
@@Babiii-g8n disagree there many professionals using it reducing pressure the meat deadlines, as long as laws are in place to protect jobs there is no issues only benefits. being disabled it helped me massively. plus as in my recent video if you want it to do a style or consistent character you need to train it requires the commission of a artist to build that data set to do so. Ai can't replace a artist its limited to its training. end of the day its only a tool and its the hands of the person using it that is responsible for the results.
its so slow even tho i have a rx6600
set memory to 6g or lower, drop step to 10 to 12, guidance to 2, and lower image size to 628x628 ish. and set gpu to MLC , stable diffusion is built on NVIDIA cuda so sadly not as good on other cards, but there are AMD users reporting faster speeds than my old GTX1070
dou didn't mention you need a kidney worth gpu to run this shit
You can run it with paid service online and use the same ip and also i run it with 6GB
What? I mentioned i'm running it on a old gtx1070, there is a guy running it on a dell with intel onboard gpu i talk about in another video on setting it up for old low hardware.
I mention the year cost of abode you can get a RTX GPU and it be one off that way rather than years of payments, also on Acly plugin page he has the setup for cheap cloud gpu renting that it works with also.
No kidney needed to be sold, just lower your size to 628x628 or 512x512 use sd1.5 lower steps and guidance and it will run on a potato
@@streamtabulous I tried to generate a 512*512 pic on my 1gb vram AMD ant it didn't render a pixel
idk
@@abdelrahmanahmed1766 yer need 4gig vram Min. 1gig would be a issue unfortunately. If you don't mind the second hand market and it's a case pc you could look at second hand gtx970, Gtx1070 both do well and are cheap.
Can get whole second hand pc with those cards cheap well cheaper that a years subscription to Adobe.
Otherwise on Acly plug it has a guide to cloud gpu and how to do that.
Apparently it's cheap like $2 a hour and you only pay for what you use. It's how most people make the models by renting cloud gpu.
I have not tried it, you could have a fiddle and see how you go.
Being the plugin and krita are free it's worth fiddling with.
That said! for the less computer minded I recommend AI playground that's free for 100 renders a day and if you put live preview on that's free, so you can save the final renders for ones your really like. And screen shot the lesser ones. To get more.
AI playground is good.
@@streamtabulous tnx for replying
I will stick now to ALPACA cloud AI processing for now unfortunately
second hand gtx1070 in my country worth a salary of two months 🫠
I think Generative fill in Krita is not 100% free , because in accounts you can see you have only 300 credits to use , it does not say if that is per month or per day , or per year. or anything about it.
there are no credits or limitations, i am not sure what you mean by 300?
krita with Ai diffusion is 100% free, no limits on generations, and no credit system.
because it runs on your local hardware
Doesn't work
read through pinned post, ensue 25gig free space when installing and start with 512x512 image size.
but read pinned posts its from users who had issues and there fixes.
Way to complicated to install on a Mac, not worth the time
LOL. There's no point in "practicing" 2D art skills any longer for 99% of the people. 😁
there is, its like saying dont bother sculpting because 3d printers can do it, there always something special about hands on. I still do some hands on dispute the pain and limitations of my disability's.
also you need artists to create images to train a model to do what you want, I trained a model on my art, I mentioned i won't copy it made something new but cool and unique
Yeah ! There is no point for a stunning walk in the beautiful mountains when you can just watch a picture of it to enjoy ! Same, right ?