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

sdxl-plusv2 seems not perform well #277

Closed Weifeng-Chen closed 5 months ago

Weifeng-Chen commented 5 months ago

I've test this version and found its performance is worse. is there any prompts to replicate the results? here's my prompts: [ 'Manga drawing of a man', 'Funko Pop of a man', 'a man with sunglasses and a corgi', 'a man in Ghibli anime style', 'a man making latte art', ] and I get: sdxl_male_0 jpg_as1 0_ss1 0_gs6_0 268

Weifeng-Chen commented 5 months ago

am I write a wrong code for inference? I modify the sd15 version so I'm not very sure..

xiaohu2015 commented 5 months ago

do you also try comfyui?

Weifeng-Chen commented 5 months ago

do you also try comfyui?

I forget the shortcut when I write the pipeline. sorry for your time.