search

SaltedSlark / Fast-3D-Talking-Face

Drive your metahuman to speak within 1 second.
18 stars 5 forks source link

Decode error exception #5

Open zhanqigithub opened 1 week ago

zhanqigithub commented 1 week ago

Follow the readme, but got error like this raise RequestsJSONDecodeError(e.msg, e.doc, e.pos) requests.exceptions.JSONDecodeError: Expecting value: line 1 column 1 (char 0)

SaltedSlark commented 1 week ago

Check server log please

zhanqigithub commented 1 week ago

Check server log please

Thanks，here is the log

File "talking_face2\lib\site-packages\torch\nn\modules\module.py", line 1130, in _call_impl return forward_call(*input, *kwargs) File "\talking_face2\lib\site-packages\transformers\models\wav2vec2\modeling_wav2vec2.py", line 857, in forward position_embeddings = self.pos_conv_embed(hidden_states) File "\talking_face2\lib\site-packages\torch\nn\modules\module.py", line 1130, in _call_impl return forward_call(input, **kwargs) File "talking_face2\lib\site-packages\transformers\models\wav2vec2\modeling_wav2vec2.py", line 385, in forward hidden_states = hidden_states.transpose(1, 2) AttributeError: 'tuple' object has no attribute 'transpose'

SaltedSlark commented 1 week ago

try to revise line 110 in wav2vec.py from

hidden_states = self.feature_projection(hidden_states)

to

hidden_states = self.feature_projection(hidden_states)[0]
zhanqigithub commented 1 week ago

Thanks

zhanqigithub commented 1 week ago

Thanks, now I can see lipmoves, but no audio data

SaltedSlark commented 1 week ago

Thanks, now I can see lipmoves, but no audio data

It is out of the project. It is about UE engine.

zhanqigithub commented 1 week ago

Thanks, now I can see lipmoves, but no audio data

It is out of the project. It is about UE engine.

Thanks

zhanqigithub commented 1 week ago

Sorry ， reopen this issue, I had tried with 30s audio, but the predict data is prediction size np arrray is 92736, 92736/32=2899 frames, which is much more longer than the audio length, can't match together, almost 3 times longer than the audio. any clue for this?

SaltedSlark commented 1 week ago

check the channel of audio file？

SaltedSlark commented 1 week ago

compare the basic information between the demo audio and your audio with the tool called sox.

zhanqigithub commented 1 week ago

check the channel of audio file？

I tried with the sample audio, get same result, the frame nums is so much longer, audio and lip movements can't match

SaltedSlark commented 1 week ago

send me the your audio data.