Using existing images

#63
by JahnX - opened

Is there a way to incorporate my existing images as part of the generated images?

Not directly.
That's a good feature request though.

We can use other tools to incorporate those such as: https://huggingface.co/spaces/diffusers/stable-diffusion-xl-inpainting
(Edit: Actually that inpainting space doesn't allow image input only masking of parts of an image to prompt a change into that location. Sorry)

how about doing lora training?

How would I go about that?

@JahnX you'd collect some images of what you want to represent and train with the dreambooth method (e.g. https://github.com/huggingface/autotrain-advanced). This gives you a "word" that's trained to represents your style/ subject.

Now you load that trained model with AI Comic factory and include your "word" into the prompt.
Check the net for "Dreambooth" to get an Idea what I'm talking about. LORA is a low ressource usage training method, Dreambooth is one Method to introduce your own subjects into stable diffsuion.

Sign up or log in to comment