THE FUTURE of FREE AI Models Is HERE! LOCAL INSTALL in 1 CLICK!

Рет қаралды 64,154

4 ай бұрын

Say goodbye to bad text-to-image Ai models and hello to the future of free open-source AI image generation: Stable Cascade! This new model from Stability Ai is taking the AI world by storm, offering for an open-source model unparalleled prompt following, clear text generation, and very aesthetic image quality for FREE. So in this video, I'll show you how to download the preview model inside its standalone webui and use it on your own computer locally for free!
What do you think of Stable Cascade? Let me know in the comments!
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
SOCIAL MEDIA LINKS!
✨ Support my work on Patreon: / aitrepreneur
⚔️ Join the Discord server: bit.ly/aitdiscord
🧠 My Second Channel THE MAKER LAIR: bit.ly/themakerlair
📧 Business Contact: theaitrepreneur@gmail.com
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
✨ PATREON LINK: / aitrepreneur
Stable Cascade: github.com/Stability-AI/Stabl...
Stable Cascade Demo: huggingface.co/spaces/multimo...
My Stable Cascade Huggingface Repo: huggingface.co/Aitrepreneur/S...
Google Colab demo: colab.research.google.com/git...
pip install torch torchvision torchaudio xformers --index-url download.pytorch.org/whl/cu118
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
►► My PC & Favorite Gear:
i9-12900K: amzn.to/3L03tLG
RTX 3090 Gigabyte Vision OC : amzn.to/40ANaue
SAMSUNG 980 PRO SSD 2TB PCIe NVMe: amzn.to/3oBR0WO
Kingston FURY Beast 64GB 3200MHz DDR4 : amzn.to/3osdZ6z
iCUE 4000X - White: amzn.to/40y9BAk
ASRock Z690 DDR4 : amzn.to/3Amcxph
Corsair RM850 - White : amzn.to/3NbXlm2
Corsair iCUE SP120 : amzn.to/43WR9nW
Noctua NH-D15 chromax.Black : amzn.to/3H7qQSa
EDUP PCIe WiFi 6E Card Bluetooth : amzn.to/40t5Lsk
Recording Gear:
Rode PodMic : amzn.to/43ZvYlm
Rode AI-1 USB Audio Interface : amzn.to/3N6ybFk
Rode WS2 Microphone Pop Filter : amzn.to/3oIo9Qw
Elgato Wave Mic Arm : amzn.to/3LosH7D
Stagg XLR Cable - Black - 6M : amzn.to/3L5Fuue
FetHead Microphone Preamp : amzn.to/41TWQ4o
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
Special thanks to Royal Emperor:
- Totoro
- TNSEE
- RG
- Judy Godvliet
- Gluthoric
- Jay
Thank you so much for your support on Patreon! You are truly a glory to behold! Your generosity is immense, and it means the world to me. Thank you for helping me keep the lights on and the content flowing. Thank you very much!
#stablediffusion #stablecascade #imagegeneration #texttoimage
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
WATCH MY MOST POPULAR VIDEOS:
RECOMMENDED WATCHING - All LLM & ChatGPT Video:
►► • CHATGPT
RECOMMENDED WATCHING - My "Tutorial" Playlist:
►► bit.ly/TuTPlaylist
Disclosure: Bear in mind that some of the links in this post are affiliate links and if you go through them to make a purchase I will earn a commission. Keep in mind that I link these companies and their products because of their quality and not because of the commission I receive from your purchases. The decision is yours, and whether or not you decide to buy something is completely up to you.

Пікірлер: 201

@Aitrepreneur 4 ай бұрын

HELLO HUMANS! Thank you for watching & do NOT forget to LIKE and SUBSCRIBE For More Ai Updates. Thx

@LouisGedo 4 ай бұрын

👋

@HeyTikal 2 ай бұрын

Il y a t il un moyen. D'utiliser en plus les lora qu on a créé pour sdxl😉

@MrMorrigan1989 3 ай бұрын

You are the first creator I came across that really does amazing work on their patreon and I created my patreon account just for that. The scripts that I got are just really nice and save me a lot of time. Amazing videos, great patreon content...

@musicandhappinessbyjo795 4 ай бұрын

The hands. It's the hands that makes it so good.

@ggoddkkiller1342 4 ай бұрын

Tears of editing fingers will finally stop..

@HeroEmiya 4 ай бұрын

The pain of generating image in 20 seconds then spending hour fixing yands

@musicandhappinessbyjo795 4 ай бұрын

@@HeroEmiya so true.

@lefourbe5596 4 ай бұрын

good short cover of cascade. i hope it will be part of the progress of stable diffusion 3. as a base model i see this as a big plus we need that X-Adapter tool to translate previous work to this newer model. (IT SHOULD COME IN 3 DAY :D)

@glenyoung1809 4 ай бұрын

Left this comment on another YT video but it's applicable here. Won't go into the aesthetics as there will be plenty of comments on that topic. This is more from a performance perspective for those who are curious. Thing to keep in mind, this is an unoptimized research grade model, not a production model so hardware requirements are sure to come down as well as an improvement in quality. Did some testing in a 4070 Ti Super 16GB Model took 13.5 GB VRAM after doing a 512x512 image. [GPU Memory reading from Task Manager] 512x512: 8.62s [13.5GB] 640x640: 9.3s [13.6GB] 768x768: 9.44s [13.6GB] 896x896: 9.6s [14.2GB] 1024x1024: 10.1 sec [14.6GB] 1152x1152: 10.5 sec [15.0GB] 1280x1280: 11.3 sec [15.5GB] 1408x1408: 13.5 sec. [15.9GB] 1536x1536: 21.8 sec (overflowed VRAM into GPU shared memory). [16.4 GB] 1664x1664: 31.6 sec [16.9GB] 1792x1792: 43.2 sec.[17.5GB] 1920x1920: 56.4 sec. [18.2GB] 2048x2048: 78 sec. [18.8GB] 2176x2176: 93-130 sec [19.2GB] (times were highly variable over 5 runs with different styles and prompts and there were visual artifacts in some images like double arms etc.)

@CraigUKgames 4 ай бұрын

how did you get it to do a 512x512 image, the sliders wont go lower than 1024 for me.

@glenyoung1809 4 ай бұрын

@@CraigUKgames I installed the ComfyUI node version which has the ability to go down to 512, 640, 768, 896.

@blacksage81 4 ай бұрын

Thank you. I'm only on 12gb and I figured that this model was going to be a bit much for my rig. I'm going to play with it once we get a safetensors version for sure, but pound for pound it looks better than the current xl base

@Heldn100 4 ай бұрын

+13gb is so much sadly I have 4070

@jamessurine3173 4 ай бұрын

@@CraigUKgamesI just typed in the number

@LuluDZulu 4 ай бұрын

love the one-click installers and the pacing or your videos. i’ve been able to achieve spectacular results thanks to you

@winkletter 4 ай бұрын

This is exciting. When I look back on how I've used SD, a lot of the better use-cases focus on generating 100s or 1000s of images with slightly randomized inputs. Only a percentage will work out, but I can still get lots of images that work. DALL-E can give me more accurate results, but not as many overall. Stable Cascade looks like in time it might be able to generate more accurate results at scale.

@TomiTom1234 4 ай бұрын

Thank you! Installed it and tested it, wow indeed is much better than SD, it does exactly what I type in the prompt. Even any Text I want it to be in the image it will be written. Only downside right now it needs lots of time to generate on low vRams, I have 3070 Ti with 8 vRAMs, and it took around 11 minutes ( 7:35 min for 20 Prior Inference Steps, and 3:20 for 10 Decoder Inference Steps ) to render one image. Well, I am sure there will be tweaks for it in the meantime.

@marcg1z28 4 ай бұрын

I have a 3070 Ti , At the end of the render I get an out of memory, Have you ever run into that issue?

@neofuturoai 4 ай бұрын

doing render in comfyui with 6gb vram in 1.5 minute. it depend of image size and compression.

@TomiTom1234 4 ай бұрын

@@marcg1z28 I wrote a replay but it is gone. Yes I faced that issue sadly. I was able to render images after installing it, but afterwards the "no enough memory" message appeared. BTW it is till now not worth it to install locally since on low vRams it takes ~10 minutes as I mentioned if the process went through. The Demo does the trick and very fast, or you can try ComfyUI with Cascade, there is one out right now.

@TomiTom1234 4 ай бұрын

@@neofuturoai Cascade is now available for A1111 , or better yet for Forge WebUi, and it needs only a minute to render 😃👍

@UltraStyle-AI 4 ай бұрын

I watched another video about Stable Cascade, and the other reviewer said it wasn't as good. But your review makes much more sense, thanks for giving me hope!

@virgilitech 4 ай бұрын

Hello @Aitrepreneur, first of all, thank you very much for all the content, I have been following your channel for a while to learn about the AI revolution, I thought you was one of the best channels for real. Putting that aside, I would like to ask if you know if is possible to use controlnet features through the API like the basic one that stable diffusion bring

@emmanuelmartin9568 3 ай бұрын

i would like to tank you for your generosity and exactitude of your info , it help me a lot to understand AI and how to use it ! 10/10

@emmanuelmartin9568 3 ай бұрын

keep it on !!!

@blizado3675 4 ай бұрын

Very interesting, it gets better and better. :D

@jimdelsol1941 4 ай бұрын

Honestly, I am impressed. Can't wait for fine-tuned models

@camilleschnakenbourg3491 4 ай бұрын

perhaps you could give use some more informations like the graphix ram it use or the velocity... I only have a 2070 with 8 Go so it probably be not enough

@Sloff1155 4 ай бұрын

This is amazing, I hope you will do another best uncensored chat bot I can only run a 7B max due to hardware limitations.

@neofuturoai 4 ай бұрын

it is already natively supported in comfyui with support of minimum 5gb vram cards with lite models

@PixelsVerwisselaar 4 ай бұрын

Dope intro 👌

@Squizz11222 4 ай бұрын

It's a nice preview, thanks. I just hope there will be ways to lower the vram since it eats around 16 which makes my poor 12 GB gpu generate one image for 5 minutes xD

@thanksfernuthin 4 ай бұрын

On top of all this goodness they say it's a lot easier to train. Imagine the possibilities!

@bobwinberry 4 ай бұрын

Fantastic! thanks

@annyeong5810 4 ай бұрын

Good video, as usual. Can't wait for the community remixes of this!

@TheSchwarzKater 4 ай бұрын

Yeah. I need a 10GB vram version... or a new GPU.

@spyknight69 4 ай бұрын

Hi for some reason when i get to the point of the requirements give me an error “Could not open requirements file: [Errno 2] No such file or directory: ‘requirements.txt’

@makulVR 4 ай бұрын

PLEASE make a Tutorial on how train a Lora/checkpoint based on this model.

@undefined6512 4 ай бұрын

Cats: "Give us the bird" Me: 🖕

@ScottTheis 4 ай бұрын

Dude, are you a mind reader? I've been looking for what comes after SDXL. Thank you for all of your content. You rock.

@shrektheogrelord6435 4 ай бұрын

safe to say this guy holds the monopoly on this content niche

@vladoportos 4 ай бұрын

Is there some text to image that can actually generate a scene ? So far all I have tried ( and I have tried many ) can mostly do single subject in close with some background... almost all of them have problems with multiple characters in a scene...

@leucome 4 ай бұрын

Multiple characters is not working that greats unless multiple characters is the main concept. By example: an "army", this is one concept that mean multiple characters. If I want some control over the scene composition, then I use the Tiled Diffusion extension for Stable Diffusion webui. It as a feature to add region with their own independent prompt. There is other extension with similar feature too. Anyway, It works to make scene with multiple characters or to make scene with a character in a specific region. But it wont be able to easily make scene with characters interacting with each other. For characters interaction then we are back to using a concept that represent the interaction. An other thing that can help with multiple characters is the sampler named "Restart". This one is a bit better at preventing concept from leaking on each other.

@cosmicclone12 4 ай бұрын

after the pip install torch completes i get an error ( ERROR: Could not find a version that satisfies the requirement xformers (from versions: none) ERROR: No matching distribution found for xformers ) does anyone know what i have to do

@chaostwizt7584 3 ай бұрын

Same error.

@LijmPistool 2 ай бұрын

samr

@chaks2432 4 ай бұрын

where are the images stored? Or do they get overwritten if you don't manually download them?

@scrup13s 4 ай бұрын

i'm so impressed by your support and sharing of all those techs, i joined your patreon! keep it up

@CrextComic 4 ай бұрын

Correct me if I'm wrong, but you'll still get far superior images with Stable Diffusion, as it enables loras and different checkpoints, which is not available for Stable Cascade (yet)?

@SergeyTsyptsynPlus 4 ай бұрын

@Aitrepreneur, Looks like the prior steps slider is broken. Makes no sense that 10 step prior and 30 step prior look abs the same. Even 1 prior step the same after editing local gradio minmax slider for steps

@MPCDesenvolvimentoWeb 4 ай бұрын

I did all the procedures for installation but always generated a random error, I tried several times following the tips, unfortunately I could not

@81HM 4 ай бұрын

If I already have dependencies like Torch installed because I have Stable Diffusion installed do I need to install them for this also?

@rodrigoruiz976 4 ай бұрын

Do you have any recommendation of a docker container I can run those models at?

@serta5727 4 ай бұрын

Quite a nice model

@TheSchwarzKater 4 ай бұрын

The most useful and reliable channel on KZfaq. Thanks once again.

@shabadooshabadoo4918 4 ай бұрын

I feel like cascade did the animals with protest signs worse. It did the dogs features better. But 3 dogs (or humans or whatever) isn't what I picture when I think of a protest.

@JUKEBOXANIMATION 4 ай бұрын

wow incredible update ! wait.. censored ? Imma wait for your next news...

@yadav-r 4 ай бұрын

Hi, can you please do a video on Upscaling SDXL generations for Metal & Canvas printing. Good day.

@GeekRedux 4 ай бұрын

Clear and direct guide for getting up and running. Unfortunately when trying to run it I get a "CUDA out of memory" error; using a 4060 TI 16GB. So I guess I need to wait for a more optimized version. (Also going to nitpick: 2:19 Those \ are backslashes, not slashes. Slashes are the ones you see in Linux and URLs: / .)

@jeffbull8781 4 ай бұрын

The research version is 20gb so although its fast and efficient it needs a big gpu to even load it.

@Galbalan 4 ай бұрын

I tested on a 4060 Ti 16GB... able to process images at default settings in 10 - 15 seconds. (Though it just BARELY fits in memory, and has gone over once. In which case I just restarted the app and continued.) If it doesn't work for you, I wonder if there's some other variables at play as well. *watching task manager's performance tab like a hawk*

@GeekRedux 4 ай бұрын

@@Galbalan Interesting. I'll play around with it some more.

@BrianBGeronimo 4 ай бұрын

Hi @Aitrepreneur, do you know an AI model for translating audio?

@QuestBboy 4 ай бұрын

I am unable to re-launch stable cascade after following your steps for the local install. Is there a batch file to launch it?

@KongLoLmf 2 ай бұрын

Did you find a solution?

@Lemankus 4 ай бұрын

The strength of the prompt response is impressive, but Immediate prompt response with a system isn't that important, especially if you can use combinations of loras to augment the process. Composition of various models will produce more innovation over time. That's the strength of a system like stable diffusion over MJ or Dall-E. If the others permit external model creation and integration to their models, then they will truly be tremendously powerful.

@m7280 2 ай бұрын

This is sickening! Thanks for the tutorial. If I install this one can I generate image2image?

@megadarthvader 4 ай бұрын

Is it compatible with your own models or loras and with control net?

@LuluDZulu 4 ай бұрын

good question. I think until thats the case its gonna be great for making a high quality base image and then zapping it with ur taste in img2img

@fenix20075 4 ай бұрын

colab error, tensor size does not match... looks like we have to try the local one....

@duskairable 4 ай бұрын

Agreed but currently it's still needs high end PC to run the model, i'll be more impressed if the preview can run on low/med end PC like SD 1.5.

@richardbond4496 4 ай бұрын

nice explanation

@siddharthmishra8283 3 ай бұрын

Awaiting for your SUPIR installation for A1111😊 12gb version

@yngeneer 4 ай бұрын

pretty pitty in doesn't store the weights in classic format somewhere in the directory structure of the demo. I saw it downloaded several files somewhere around 20gb, yet the app folder nor the env folder does have it in it. It is stored somewhere in huggingface cache on the system drive which is not suitable for me, i need to have it all stored on another drive... :(

@gersonhiroshiyoshinarijuni3846 4 ай бұрын

thanks!

@Vitautas8 4 ай бұрын

I did it, it worked, how to relaunch it?

@tetsuooshima832 4 ай бұрын

Yeah I tried that a few days ago, not too convinced coz it's slow on my 3060 AND it's censored so it's a complete No for me haha yet another model that will take time for the community to improve, SDXL took more than 6 months to get decent anime models because not everyone can train it to begin with, so I wonder about this one... Wait and see (oh dang the video didn't start yet I'm already commenting haha)

@Aitrepreneur 4 ай бұрын

SDXL was exactly the same when it was released, once again it's up to the community to really make them better to what we need, stability ai can't because of regulation and potential lawsuit this could bring but let me tell you that once we get our dirty paws on that final model, with better prompt following and good finetuning....this will be insane...

@ramsescamara7334 4 ай бұрын

The quality is really impressive, but on my 8GB 4060 it took about 20 minutes to generate an image with 100% usage (32GB of RAM at 3200MHz)

@mehmetsahin-5 4 ай бұрын

Hello. Where are the two *.bat files, where you drag one into the folder so that it works automatically.

@chefaku 4 ай бұрын

9:31 that prompt poor Sakura xd

@minionminion6480 4 ай бұрын

can you do stuff like in sdxl like having the same model (person) in different captions

@vdbyUA 3 ай бұрын

крууууууто!!!! Тепер я стану дуже багатим ))

@AI_Image_Master 4 ай бұрын

I tried it on hugging face and it certainly did not do what I wanted in the prompt. In my test I wanted a photographer shooting a model at a high fashion shoot with basically the model on one side and the photographer on the other , you get the idea, and it did every thing but that. I even got a photographer pointing the camera backwards and looking into the lens. Could not get what I wanted. Also it is censored. As a test I asked for a simple nude portrait. And worse then telling me that it can not do that it gave me a fully clothed model every time. Another test of 2 women dancing gets distorted faces and bad hands. 1 women OK'ish but not the pretty face that I asked for. It seems that it does cats fine. I tested that. Human faces not so good. I am not going to install locally since from the other comments I certainly do not have enough vram. But SD does run fine and is quick locally. And yes SD does not do the Photographer and the Model image that I want either. Still need to do a bunch of inpainting and other tricks to get the exact shot that I want.

@shabadooshabadoo4918 4 ай бұрын

mines taking like uh over an hour to do one image. I think because CUDA is enabled? Any way to disable it?

@SaltyMaud 4 ай бұрын

How well it can be finetuned and how well it will be supported by tools like controlnet is ultimately what decides the fate of this promising model.

@TheSchwarzKater 4 ай бұрын

It can be trained much faster than SDXL. Controlnet and other models will follow, I don't see why they wouldn't.

@foreropa 3 ай бұрын

As Always, this doesn´t work with AMD cards, or does it? I tried one option of another video, and it said I didn´t have the Nvidia drivers when installing it.

@randymonteith1660 4 ай бұрын

Its fast with my 3090 24 Megs card . Takes about 30 minutes to install using your " Install " file

@luqiufu 4 ай бұрын

It stops using "SD" name! Is it different from sd model foundamentally? Or it is way better than all sd model?

@KimiMorgam 4 ай бұрын

you forget the command line on description

@borisrusev9474 4 ай бұрын

Now try to make it generate full-body images. Or images which are not portrait shots where the subject is directily looking at the camera on a blurred background.

@chiragforu 3 ай бұрын

can i train my own dataset and run command on cascade?

@carnajom8831 Ай бұрын

Don't suppose anyone can help me, I receive the following errors when run the pip torchvision line "ERROR: Could not find a version that satisfies the requirement xformers (from versions: none)" and then "ERROR: No matching distribution found for xformers"

@n0b0dy_know 4 ай бұрын

Amazing KZfaq Channel

@PlepperGuy 2 ай бұрын

When installing python make sure to checkmark add to PATH in installer

@banzai316 4 ай бұрын

Thanks, Aitrepreneur 👏🐐

@Aitrepreneur 4 ай бұрын

You bet

@fr0zen1isshadowbanned99 4 ай бұрын

yay :D bump

@meyou7041 4 ай бұрын

this is on civit, can't i just download it from there

@AmirZaimMohdZaini 4 ай бұрын

All this thing, it still requires 20GB++ GPU which is Nvidia Tesla Maxwell architecture and etc. Might plan to buy this kind of GPU solely for quicker SDXL Lora training.

@7satsu 4 ай бұрын

Who do my downloads in cmd always stop half way thru and just cancel entirely .-. it took me 20 attempts to finally install the requirements and now I can't download even half the 7GB safetensors file because the download gets interrupted in cmd every single time, even with Ethernet connection

@kilosera 4 ай бұрын

Unplug your modem and router from power, let it cool down, replug everything back in and then try again. I had the same issues with spectrum internet back in US. Everything from python repos to llms repos just dropped randomly. It's your connection, unfortunately.

@7satsu 4 ай бұрын

@@kilosera much appreciated, i'm legit on Spectrum too lol. Well that's frustrating, that's why I try to download the files I need within a web browser because at least it will let you resume the download

@Woolfio 4 ай бұрын

I assume once again this does not work for AMD graphics cards?

@Jonathan-gd7qt 4 ай бұрын

I wish AMD would build compatibility for all of these models.

@MEGTON 4 ай бұрын

How to download the model??

@rufflesal 2 ай бұрын

Can the images be used commercially?

@celozzip 4 ай бұрын

i just had a look in stability matrix... 34gigs....

@TweedReborn 4 ай бұрын

please excuse my simple-mindedness but how do you relaunch this particular branch after you have closed it?

@Aitrepreneur 4 ай бұрын

open a command prompt window inside the stable-cascade folder, activate the environment env\Scripts\activate then python app.py (followed by an argument)

@shabadooshabadoo4918 4 ай бұрын

lmao. Useless Sakura. You nailed it brother!

@darkmax2047 4 ай бұрын

How big is the download size?

@Steamrick 4 ай бұрын

Very nice. Too bad it pegs my GPU at 20GB VRAM usage. That's gonna take a lot of optimization.

@AIEnthusiast199 4 ай бұрын

is it uncensored?

@procrastinator1727 4 ай бұрын

Can this model work with Automatic 1111?

@Drakmour666 4 ай бұрын

How much of Vram needed for the model to work on local PC?

@Steamrick 4 ай бұрын

20GB, as mentioned by Stability AI

@rinpsantos 4 ай бұрын

My Stable Cascade extension for A1111-Forge works like a charm on my RTX4070 12GB VRAM.

@Drakmour666 4 ай бұрын

@@SteamrickAh, so again hungry pup. :-D

@DjordjePivnicki 4 ай бұрын

For me it works with 6GB in ComfyUI.

@AmirZaimMohdZaini 4 ай бұрын

20GB GPU requirements which is Nvidia Tesla Maxwell, Pascal or Turing model would be suitable for this. Also, these cards are now dirt cheap today.

@PokemonButcher 4 ай бұрын

How does one open the instance again once it has been closed? 😅 Thank you, kindly!

@Aitrepreneur 4 ай бұрын

open a command prompt window inside the stable-cascade folder, activate the environment env\Scripts\activate then python app.py (followed by an argument)

@omkarjamdar4076 3 ай бұрын

Can these dowloaded huggingface llms work offline?

@TheCyberMacGyver 4 ай бұрын

Hi, how can I run this on Linux?

@AIKnowledge2Go 4 ай бұрын

While Stable Cascade is revolutionizing the AI image generation scene with its open-source model, i am not sure the stable communtiy follows. SD 1.5 is still the primary used Model on CivitAI. To me Stable Cascade is promising a significant leap forward for content creators and AI enthusiasts alike. 🌟🎨

@jojrab5207 2 ай бұрын

When running the env\Scripts\activate command I receive "The system cannot find the path specified"

@PlepperGuy 2 ай бұрын

When installing python make sure to checkmark add to PATH in installer

@geffridamer9798 4 ай бұрын

sadly doesn't work on colab

@j.jarvis7460 19 күн бұрын

ERROR: Could not find a version that satisfies the requirement xformers (from versions: none) ERROR: No matching distribution found for xformers anyone know what happend?

@ag36015 4 ай бұрын

is this heavy on GPU as other image generators?

@neofuturoai 4 ай бұрын

its more like SDXL, but it is generating in 2 stages with 2 models (more like 4 models with text encoder and vae) . you can run it with minimum 5gb vram in comfyui already