usage in diffusers

#2
by onlyoneplease - opened

I tried using this lora to remove a car from an image but I keep getting a smudge where the car was. is this normal?

image.png

Yes, it is normal because this lora trained on very small dataset and doesn't handle all remval task good enough

I see, It's still better than flux-fill
I'm really interested in this lora so if you could share the training script or I could supply ~50 image pairs of car images before and after removal to add to training that would be great!
I was thinking of fine tuning https://huggingface.co/sayakpaul/FLUX.1-dev-edit-v0 but I'm not sure it'll give good results.

I open sourced the training script at https://github.com/lrzjason/T2ITrainer

Sign up or log in to comment