Stable Diffusion | FORGE is BETTER than Automatic1111 | ONLY 4GB VRAM Needed! | Step-by-Step Install
HTML-код
- Опубликовано: 29 фев 2024
- FORGE is a Stable Diffusion interface that is built upon the popular Automatic 1111 user interface. But unlike the standard Automatic 1111, FORGE offers an increase in performance while reducing the system resources required to generate images, allowing Nvidia users with as little as 4 GB of VRAM to finally make use of Stable Diffusion.
But it's not just for low GB GPUs. The speed and performance benefits can be seen in all GPUs, making FORGE a truly great way to create images using Stable Diffusion.
And the installation is fast and easy!
Links in the video:
▼ #1 - github.com/lllyasviel/stable-...
Thank you for watching! ♥ - Хобби
saw this in my feed, I do Stable Diffusion Videos. I found an epic flaw with Forge. It can't generate images with wide aspect ratios but Automatic1111 can.
This is some important and much needed information since this could impact a large number of potential Forge users - thank you!
thanks heisenberg 🧑🍳
We're gonna cook with this one
Great, Thank you.
Thanks! Best video on installation and also on a version that will work on my old card.
Thank you - Glad it helped
Any way to get wildcards to work on forge? My extensions has it listed but doesnt actually show the tab on the front UI
Thank you for this. I had been pulling my hair out all day trying to get SD installed follwing another persons guide (think it may have been outdated), having all sorts of errors with pytorch installations and webui batch load issue. Eventually found your video and I was installed in a matter of minutes :)
Thanks, is impresive fast. how i can add or change the woman face factions i mean for example from a picture? thanks
any suggestion for sdxl? for now it takes about 5 min for one photo
Yes, i can confirm that it works great with gtx1050ti 4gb. But my other pc is ryzen7735 with dedicated 4gb vram and no chance getting it to work.
Can I use comfy ui with 4gb nvidia vram
forge starts extremely slowly for me or not creating images at all is painfully slow. I have a gtx1080ti with 11gb and fooocus and comfy work great
Thanks. I've never used a "detailer."
Actually, it'll run on 2GB also at about 10s/iteration.
It helps immensely that it supports SD 1.5 models, of which there are many pruned versions that can be loaded almost entirely into the GPU.
all gpus? 4gb? now my rx 580 can do something
Is there a code link for Kaggle?
hi is there a way to get dark mode with forge i have it with automatic 1111
?__theme=dark
I am using rtx2060, and Forge solves some speed problems when i try to use SDXL models... Its much much faster.. When i use 111 it can take 30-1hr for 1 render... vs Forge which might take 5-10min or lesser. The good thing is that i have both installed.. and they both share the models.. so if i really need i can just switch back to Automatic1111 ... so far i have not really needed to do that..
RuntimeError: Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check
Same here, even though I have 8GB vram ...but its AMD : (
When you try to generate a video of walter white but forget to use control net
No Linux tho sad
Dicoba tapi tak bisa jalan
Can I run it on Intel HD Graphics?
Unlikely. You'd need at least 4gb NVIDIA GTX. Maybe even 2GB, depending.
Your video feed, the talking head portion, is overly processed and sharpened.
It's quite uncanny, maybe there's a way to tone it down?
What he doesn't mention. Don't use it if you are on anything except a Nvidia card. Radeon cards will have to run --directml but will probably have issues.
Jesse we need to cook
What does this give us that Fooocus doesn't?
doesn't fooocus force you to use SDXL model? It's slow as heck on my laptop
The main thing Foooocus has going for it is the simplified UI. A lot of people are intimidated by Automatic1111 and especially ComfyUI. And FORGE features an even more complicated UI than base A1111. But with that complexity comes greater control where you are able to more precisely define the desired output and therefore more closely generate a specific desired image.
@@011egis basically yes, you 'can' use SD 1.5 models as refiners but you don't need to use a refiner with most XL models. So basically yes it's XL only.
@@ronpiersonhay man i just installed on my laptop with RTX 3050 4gb Gpu, tried couple of times, suddenly laptop shut off, what happened to my pc? it's not turning on at all. tried 40 times after cooling down, what'd i do now? I'm gonna die, what happened? 😔😔
using windows, sorry would done vote but the tutorial is good so...
Issues occur with FORGE when updating Pytorch and CUDA. For FORGE you need to do a downgrade.
Like always, just for Nvidia.
So the "all GPU's" mean all Nvidia GPU's.
Come on AMD pick up the slack, I also want to play with this.
It takes a lot of jumping through hoops to make it work on AMD, and lots of googling crap, but I have absolutely made it work. Your options are to get an NVIDIA GPU or put in the time and effort to get SD1.5 running following a mix of old guides. Lots of them say they apply to AMD but only apply to 6000 series + and lots of them say it also works for AMD but then go "hurr durr but you'll render with the CPU and it'll take a week" so it is absolutely awful, and by the time I was done it I don't even know what kind of hell I sifted through but I put it all behind me and will probably jsut give up on AMD after this card.
Fewer resources - for god's sake.
Dark Mode sir!
?__theme=dark
you look like you were ai generated
It might be faster, not better. The quality is really poor next to a1111, at least when it comes to img2img and batch sequencing. I’m very disappointed…..same model, same seed, same everything and still getting completely different results…..same as comfy…..