Stable Diffusion on Windows: How to Install Visions of Chaos
Вставка
- Опубліковано 22 сер 2022
- First video in the Visions of Chaos Series.
Download Visions of Chaos: softology.pro/voc.htm
Support Visions of Chaos: / softology
More to come soon. Let me know in the comments what you would like to see next!
Visons of Chaos is an amazing bit of software, and I love that SD is included. If you figure out to run Dreambooth training in VOC, that would be a really awesome tutorial video!
So glad I found your channel! I can't wait to see how i can mix this with my design!
This is fantastic, thanks so much mate!
Thats amazing
Keep it up bro
Appreciate you
Many thanks, this gave me that extra encouragement and instructions to get it working!
Very informative. Thanks bro.
Pro-tip: If Visions of Chaos says you didn't install CUDA even though you followed the instructions perfectly, double check that you don't also have the latest version installed. If you have v11.3 installed from some previous project, then VoC will try to use it instead of v11.2 and it won't work. Simply uninstall v11.3 and any of its components through Add or Remove Programs. This resolved the issue for me.
Same !
is this still the case 1 month later?
@@Coltography I don't think anything has changed. Basically just follow the current VoC install instructions on the website and when it comes to CUDA, make sure you install whatever version it asks you to install, then check if you have a newer version already installed. In the CUDA install folder I had two folders. One marked v11.2 and one v11.3. I ran the uninstaller for v11.3 (maybe just deleted folder? Can't remember) then VoC worked.
@@Coltography to clarify, don't get too caught up on the specific version numbers I mention. Just install the version they ask you to in the install instructions and make sure you don't have a newer one also installed.
Real cool animation bro
Thx for your help too !
I'm gonna walk in here with another question
every time I have Vision of Darkness generate an image, even if it is the same prompt and seed, it reloads the model
which takes significantly longer than actually rendering the image
is there some way to keep the model loaded?
Thank you! I had my environment all set up I just couldn't find stable diffusion in the machine learning options!
great tutorial - thanks! do you know how to train a model in it? like we do in dreambooth?
How can i set how long or short the animations are? I’m kinda confused on the output part
Hi! Thanks a lot for the video. All works fine. Have only one issue - I made all of the setups on SSD Z:\....., but all of the packages are downloading to C:\Users\.....\AppData\Roaming\Visions of Chaos\.... My disk C doesn't have enough space, how to fix this?
Does having two graphics cards free up memory because one is not being used as a display
Where do you set the movie length? IE Frame count? You just let it run?
Also could you do a tutorial on the deforum SD video settings please? I change the defaults and it still renders out universe default.
Awesome! Thank You :)
this is genius, thanks
i want to know how to deepdream never got a chance to do it locally
Is there any advantage to running sd in voc vs just locally with just lstein ver or something?
I am not 100% sure what is possible within the other methods you are talking about but, I will say there are a ton of advantages of having VOC as a creator, maybe I will make my next video on this. Aside from all the other tools and scripts included with VOC that can create depth maps, upscale, interpolate frames, improve faces, and so much more, not to mention the 30+ other text to image scripts. When it comes to Stable Diffusion, you can batch process the same or a list of prompts, utilize initi images, create animations, I also believe the dev is working on getting inpainting and other features from various stable branches added.
Win11/rx 6800 how to run the Webui on locally ip
Hey, it's me again.. with questions ! How to resume the run of an animation on disco diffusion ? just checking the "resume_run" box don't work
I have never used the resume_run function. I try to remember to save my TTI files for animations and then if I ever want to continue it, I will run a new animation with the last frame selected as an initi image.
So you have to have a NVIDIA card then? You can't run this with an AMD card?
took me a while but i got it installed thanks to you... but when i tried running stable diffusion it doesn't have any models? Any idea. again this was very helpful.
hmm... it should have downloaded it. I see the most recent version of VOC lets the user choose what scripts get downloaded, unsure how that works yet. If you try generating an image, it might download what you need on its own, if not you can find the models here: huggingface.co/CompVis/stable-diffusion-v-1-4-original/tree/main and manually place it in \Visions of Chaos\MachineLearning\Text To Image\Stable Diffusion\models\ldm\stable-diffusion-v1\sd-v1-4.ckpt
@@doctordiffusion thank you. i found it now. that was helpful.
are there any hopes for rtx 2060 ? 6gb vram
I do not think, at the moment, There will be more models released that should require less VRAM in the future.
does this need a particularly fast drive, or can you put it on a HDD and let it spin?
Will run faster on SSDs but they are not required
@@doctordiffusion
Wrong question….
I have a PCIE SSD and a WD red at 5400 rpm
One reads at 1.5Gb, the other at 120Mb
How big would be the benefit of having it on the SSD?
My animations are definitely not like yours and I’m not speaking about the upscaling, etc. Everything seems to fall apart after about 10 frames or so. Like there’s some kind of auto sharpening going on that just fries the image, no sharpening variables are applied by the way. I noticed I did not have the clip features they were grayed out in recursive settings. any help would be appreciated, plus a tutorial on your upscaling workflow for animation would be excellent!
I wish I saved more of my settings, but I am working on a settings overview video.
I believe mine is doing the same thing. I had pretty much all the same settings as we set in the video except I was using the default prompt and seed as a test
I think I figured it out. If you uncheck "random seed each frame" then it uses the same seed every time. This causes issues as it tries to create the same image over and over while zooming. Looks a lot like sharpening.
Will this work with an AMD GPU? Specifically, an RX 6700XT with 12GB of VRAM? Everything I've seen/read so far says "NO!". I would really like to start using Stable Diffusion locally but the fact that I went Team Red has prevented it. Thanks.
No.
You need cuda cores.
Just switch to Nvidia. There's no point on using AMD cards
@@aeit999 progress now ? Cant get another since this is my First since 10 years ....
Just 400gb of space and 10gb of gpu ram...
i have 8043Mb of vRAM is it ok ?
I have a rtx 3070 so i am asking myself if is it ok because it's written a least 3080
@@unifio8130 I have seen some people report success with only 8 GB and getting Stable Diffusion to work. I only have the one GPU at the moment, so I can't test this on my end. Even if Stable won’t work, there are a lot of other tools and scripts within VOC. If you end up trying it out, please let me know if it works.
@@doctordiffusion ok thanks for the answer, i will download everything this night and i tell you that later !
@@doctordiffusion i did a comment that give everything about that, check that :)
@@doctordiffusion also i have some questions now :
Can we make a video with stable diffusion ? or it's only with disco diffusion
What is you VoC theme ? :)
What is your ssd speed just to know if my ssd that i lost somewhere is a thing to refind so faster generation hehe
Anyone else's art becoming progressively more monochrome the longer it runs?
You need 400gb?! Holy shit I don't even have 100, and not even uninstalling Final Fantasy 7 Remake will make up for it.
Right after I made this video the Dev changed some things and I think it only downloads what you use.
don't laugh at me please, did anyone tried with a 1070 8GB?
i have the same graphic card and i will try tomorroy so if you want i will give you info
i made the comment now, you can check that !
Yes please. Does this work with you?
Yes it worked ! I said all the benchmark and things you can do in my comment to that video (search a little bit) you're limited to low-medium resolution depending on the stable/disco diffusion
currently installing it
it... needs a lot of things
okkkk to answer to @worstimever and @SYNTIENCE annnnd, yeah this work with an RTX 3070 (8Gb of Vram)
So here some benchmark (i also have 30Mb/sec HDD and the program use a lot the hdd before batching so you win times if you have great ssd) :
I started with Discodiffusion :
with 80 iteration and 778x778 work fine (8 mins/img)
same with 448x448 res and it took 5 min/img (that takes around 6Gb Vram for who asking)
I did the same with stable diffusion now :
with my hdd it tooks a little bit more time to prepare the generation (around 3-4 minutes) it slows a lot the process sadly
50 iterations with 778x778 is impossible to do
same with 448x448, it takes up to 7.3 Gb of Vram, with 10 sec/img (very fast wow) but i have 4 minutes to wait before because of my hdd again so, yeah
Also note : you need at least 15Gb of ram, it takes a lot of ram when generating image
Hope i help anyone who read this, if you have questions about that you can ask me in the answer section
Thank you so much for sharing this information!
Holy shit. Ok that's what's wrong. Mine is super slow. I was worried about the 400gb requirement so I put it on an HDD. Switching to my SSD later tonight
Also the fact that it has to load the model every time seems unnecessary.
@@NormanTiner VOC update allowing batches to run without reloading model is coming soon.
@@doctordiffusion awesome. Yeah that would cut the time down significantly. Like a longer video this could save hours.