search

BadToBest / EchoMimic

Lifelike Audio-Driven Portrait Animations through Editable Landmark Conditioning
https://badtobest.github.io/echomimic.html
Apache License 2.0
2.36k stars 274 forks source link

为什么face_locator_tensor 这么慢呀。。。69s一步 #101

Open 3682483 opened 1 month ago

3682483 commented 1 month ago

To create a public link, set share=True in launch(). video in 24 FPS, audio idx in 50FPS whisper_chunks: (361, 50, 384) audio_fea_final: torch.Size([1, 361, 50, 384]) ref_image_latents shape: torch.Size([1, 4, 64, 64]) face_mask_tensor shape: torch.Size([1, 1, 1, 512, 512]) face_locator_tensor shape: torch.Size([2, 320, 1, 64, 64]) 3%|████ | 1/30 [01:09<33:30, 69.32s/it]

leegang commented 1 month ago

我也有相同的问题 特别慢，感觉 GPU 没被有效利用

但我不确定是不是 ComfyUI 插件带来的问题

https://github.com/smthemex/ComfyUI_EchoMimic

zjulkw commented 1 month ago

Same question

luyuhua commented 4 weeks ago

我也有相同的问题 特别慢，感觉 GPU 没被有效利用

但我不确定是不是 ComfyUI 插件带来的问题

https://github.com/smthemex/ComfyUI_EchoMimic

应该不是comfyUI问题，我用项目自带gradio启，也是同样问题 112s/it

Arvrairobo commented 4 weeks ago

it seems like project is dead, cause no one is answering

3682483 commented 3 weeks ago

on no..everybody resolve it?