- Aug 4, 2020
- 2,383
- 3,049
you need to generate smaller images. Best I could get with 4GB was around 512x512
- Dec 31, 2016
- 316
- 708
would have been a lot easier if you'd actually said that from the start considering there's nothing in the error you pasted that suggest it.
What card is also rather relevant considering things work better/easier with nvidia.
Without knowing anything about your system, there's a possibility that there's an onboard gpu being used and not the actual graphics card, the card can be unsupported, or you might have not set it up correctly so the config isn't even set to use the card
- Jan 20, 2022
- 200
- 1,954
It says you have a 4GB memory GPU
check your task manager and confirm it
If true, try to generate a low resolution image,
- Aug 4, 2020
- 2,383
- 3,049
check that you're actually using that card for stable diffusion and not some integrated graphics. Idk how, check GPU usage in MSi afterburner when stable diffusion is running or something
sick design, actually.
Kinda offputting that this GPU has...are those USB ports?)) AI's are awesome
Lol yeah
Well part of that is trying to train the duel text encoders CLIP-ViT/L & Open CLIP-ViT/G with a program that doesn't recommend it...With a new concept (GPU and graphics card prompts will draw motherboards, and all kinds of stuff in SD XL)
The other part was my training was brute forced with small latents at a very high learning rate, with one epoch
Last edited:
- Dec 23, 2018
- 481
- 1,999
Does the gfx card have a lot of memory, or does the system. Remember there is ram and vram - you might have 16gig of ram and only 4gig of vram. What model of gfx card do you have?
Here is a .png for a quick photorealisitc setup in comfy for GodOfPandas Lora for
(If your new to comfy just load the .png like a workflow)
You must be registered to see the links
(If your new to comfy just load the .png like a workflow)
Intel claims training at Full FP32 @ 467.7 Images a second with 116 Batch size on a $1600 dollar processor
I'm not sure what the latent size was but Intel is known for some low key amazing tech.
Disney used Intel to do all the graphics for Moana, those 3d models where in the terabytes and that is almost 10 years ago.
I've used AMD, Intel, ATI (Before the demise) NVidia products...I'll use what ever works.
I'm not sure what the latent size was but Intel is known for some low key amazing tech.
Disney used Intel to do all the graphics for Moana, those 3d models where in the terabytes and that is almost 10 years ago.
You must be registered to see the links
Acceleration
You must be registered to see the links
(Where I got the info from)I've used AMD, Intel, ATI (Before the demise) NVidia products...I'll use what ever works.
- Dec 21, 2020
- 1,523
- 3,589
1-to-3 image ratios seem to naturally create full-height renders. This is at least 3-5x faster than using OpenPose ControlNet. Then, if needed use outpaint to widen up the image.
You don't have permission to view the spoiler content.
Log in or register now.
Last edited:
any good prompt to do lift up/roll up skirt/dress better
anytime I do with controllnet
it tends to pulldown or have weird shape like panties around pussy
anytime I do with controllnet
it tends to pulldown or have weird shape like panties around pussy
- Dec 23, 2018
- 481
- 1,999
I believe it was quite hard to prompt for that, which is whythere are a few loras
You must be registered to see the links
for instancethanks I already try that
but it not work because I don't actually want skirt lift pose
I want to prevent it in sex position with cloth
- Jan 24, 2020
- 1,401
- 3,793
If you want to prevent it then maybe try to use them Lora's in the negative prompt. Just a thought, since you are not being clear about what you are actually looking for. When you want something very specific it's probably best to either train a Lora yourself or use controlnet.
- Jan 24, 2020
- 1,401
- 3,793
You must be registered to see the links
It's about fixing or controlling colorbleeding with regional prompting, for anyone who are interested.
What combination of model, vae and Lora, do you recomend to make an image where woman and boy are together? To make something
Because i have a problem, with "anime models", i can´t make males, only woman.
You must be registered to see the links
Because i have a problem, with "anime models", i can´t make males, only woman.
- Dec 21, 2020
- 1,523
- 3,589
Teachers cosplay imperial stormtroopers. Kinda made me think Jango Fett was the wrong guy to be cloned.
You don't have permission to view the spoiler content.
Log in or register now.
- Dec 21, 2020
- 1,523
- 3,589
I doubt. I think faces, nipples and hands have almost the same root cause which transcends the models and is rather a part of the methodology. I.e. the SD models actually create crappy faces - we do use face restore modules to fix the faces. Hence, in theory there should be a nipple restore module to make nipples look natural.
Last edited: