training to enhance the two people capability

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

5.08k stars 331 forks source link

training to enhance the two people capability #183

Open syusukexrj opened 9 months ago

syusukexrj commented 9 months ago

I want to enhance the capability of generating two-people picture and faithfully restore everyone's face id, clothing, glasses, hat and etc. Any suggestion about data preparation? Is it possible to achieve this by ip-adapter?

xiaohu2015 commented 9 months ago

for two person, I think you can use attention_mask https://github.com/cubiq/ComfyUI_IPAdapter_plus?tab=readme-ov-file#attention-masking

whiterose199187 commented 9 months ago

Hello,

Is there a way to do this without using comfy ?