If you like my work and would like to help me out follow my new X account
https://x.com/fitcorder_ai
If you want to generate an I2V you can start here
If you wish to contact me you can at https://fitcorder.com
the videos in the wan series was generated from this workflow https://civarchive.com/models/2048404?modelVersionId=2321009
I trained the Wan2.2 loras on my 4090 and the fidelity could be higher if i had a rented gpu to go higher. If someone is interested in funding that contact me on my website through the email form.
This LoRA was trained on the Flux/Pony base to emphasize a very specific visual target—full, round breasts and a slim waist—while maintaining clean anatomy and photorealism. It works best when paired with the Flux Perfect Base or base Flux.1 Dev checkpoint to enhance consistency and detail.
Please share what you create! User submissions rock.
20k buzz [AWARDED] Bounty here if you want to compete to help influence my future samples.
Videos were made with text 2 image using the bouncing walk WAN linked in suggested.
Feel free to Comment or DM me if you have any questions or want to collaborate.
Description
FAQ
Comments (13)
Is this compatible with NF4?
I donot think so
@happyhen Got it, thanks.
RIP :(
https://www.reddit.com/r/StableDiffusion/comments/1eqssqd/running_a_lora_using_nf4_flux_checkpoint_in/
Doesn't seem NF4 supports lora
@fitCorder It does now! please make a NF4 version.
cuda out of memory, 16GB Vram ^^
@SomebodyOnceToldMeh Where did you hear NF4 supporting Loras? I cannot find any news about it and have been looking into quantizing existing loras to NF4
Maybe I don't know what I'm talking about, but it may. If you are using ForgeUI, At the top it will say "Diffusion in low bits", so set that to "bnb nf4 (f16 LoRA)".
God approves this Lora although some of his followers might be shocked. God himself is tolerant and full of love.
Good God almighty, people need to hear this!
This is one of the most used lora model all over the internet.
Works incredibly well! Thank you!