search

Stability-AI / stability-sdk

SDK for interacting with stability.ai APIs (e.g. stable diffusion inference)
https://platform.stability.ai/
MIT License
2.42k stars 336 forks source link

How could I reproduce results in dreamstudio using an image + a prompt with the APIs ? #194

Closed ghw604 closed 1 year ago

ghw604 commented 1 year ago

Hi, I want to edit an image with a prompt. In dreamstudio, the result is good. But using this sdk, the results are worse. Input image: image text prompt: a photo of noodles with a fried egg 1) dreamstudio params: image Results: 4081504159_a_photo_of_noodles_with_a_fried_egg

2) SDK params: python -m stability_sdk generate "a photo of noodles with a fried egg" -i "../NiuRou4.png" --start_schedule 0.5 --end_schedule 0.05 -W 512 -H 512 --engine "stable-diffusion-512-v2-0" Results: generation_generate a photo of noodles with a fried egg_1678786996_0 If models in the two are the same? How could I reproduce the same result? I would appreciate it if you could help me.

ghw604 commented 1 year ago

I've found the difference. Dreamstudio may resize the input image to 512 x 512. When resizing image to 512 x 512 before calling SDK, results are the same.