search

tencent-ailab / IP-Adapter

The image prompt adapter is designed to enable a pretrained text-to-image diffusion model to generate images with image prompt.
Apache License 2.0
4.5k stars 296 forks source link

Stronger sensitivity to facial expressions #252

Open amelio-vazquez-reina opened 5 months ago

amelio-vazquez-reina commented 5 months ago

I'm using v2 plus with scale 0. I'm wondering if there is a way to make generation respond more aggressively to facial expressions.

I assume scale 0 is the lowest value.. Any other way to follow it expressions in the prompt more directly?

Example prompt:

prompt =  "Man laughing a lot. Wearing a black sweater, solid gray background. Only show the face. closed eyes. Looking left."
xiaohu2015 commented 5 months ago

Although it is currently editable to a certain extent, it still has limitations. We will continue to improve it later.