Fine tuning with new subjects
Advaya24 opened this issue · 1 comments
Hi,
Congratulations on the project, this work is truly exciting!
I have a question about fine tuning: will it be possible to fine tune your model with my own images and prompts? Do you have a training script I can use (similar to what dreambooth does).
Another question is: do you have a variant of this in the image to image space as well (stable diffusion)?
Hi @Advaya24, thanks for your interest!
Our method does not require fine-tuning but rather operates on-the-fly on an existing Stable Diffusion model.
You could apply our method on a version of Stable Diffusion fine-tuned with Dreambooth, could be very interesting to see the results :)
All you need to do is apply our pipeline to the trained weights of your version of Stable Diffusion.