BadToBest / EchoMimic

Lifelike Audio-Driven Portrait Animations through Editable Landmark Conditioning
https://badtobest.github.io/echomimic.html
Apache License 2.0
2.26k stars 263 forks source link

如果video_length超过face_mask_tensor的长度会崩溃 #93

Closed XeoOuYang closed 1 month ago

XeoOuYang commented 1 month ago

infra_ audio2vid acc. py中,如果video_length超过face_mask_tensor的长度,会报错。因为在推理过程中pose_latents_cond = torch.cat([face_locator_tensor[:, :, c] for c in new_context]).to(device)的c会越界,从而导致崩溃。

XeoOuYang commented 1 month ago

acc与之前区别是arg.L默认1200,而原来是240的差异

JoeFannie commented 1 month ago

多谢,这个问题在后续的版本会修复