How can we use high-quality modes?

@mnbv7758 Thanks for your interest!

if you want to try high-quality mode, please consider:

use the saved HD video in line 369 in app.py
replace line 370 - line 374 with a HD lipsync model, in our HD demo we use in-lab model. for open-source version, I suggest you take a look on wav2lip-hd

Can using wav2lip-hd solve the "frame skipping" problem?

@mnbv7758 frame skipping is caused by threshold to create the graph in

Lines 232 to 235 in e24a2f4

    
           if trans_similarity < threshold_trans: 
        
               if np.sum(position_similarity < threshold_position) >= 45 and np.sum(velocity_similarity < threshold_velocity) >= 45: 
        
                   graph.add_edge(i, j, is_continue=0)

you can make the threshold be higher

I replaced the model with wav2lip384 and it reported the following error:

Moviepy - Done !
Moviepy - video ready ./outputs/gradio/test_0/retrieved_motions_0/audio_0_retri_0.mp4
Using cuda for inference.
Load checkpoint from: ./Wav2Lip/checkpoints/checkpoint_step000318000.pth
Traceback (most recent call last):
File "/AI/TANGO/./Wav2Lip/inference.py", line 322, in
do_load(args.checkpoint_path)
File "/AI/TANGO/./Wav2Lip/inference.py", line 294, in do_load
model = load_model(checkpoint_path)
File "/AI/TANGO/./Wav2Lip/inference.py", line 167, in load_model
model.load_state_dict(new_s)
File "/AI/miniconda3/envs/TANGO/lib/python3.9/site-packages/torch/nn/modules/module.py", line 2041, in load_state_dict
raise RuntimeError('Error(s) in loading state_dict for {}:\n\t{}'.format(
RuntimeError: Error(s) in loading state_dict for Wav2Lip:
Missing key(s) in state_dict: "face_encoder_blocks.2.3.conv_block.0.weight", "face_encoder_blocks.2.3.conv_block.0.bias", "face_encoder_blocks.2.3.conv_block.1.weight", Unexpected key(s) in state_dict: "audio_refine.0.conv_block.0.weight", "audio_refine.0.conv_block.0.bias", "audio_refine.0.conv_block.1.weight", "audio_refine.0.conv_bl size mismatch for face_encoder_blocks.0.0.conv_block.0.weight: copying a param with shape torch.Size([8, 6, 7, 7]) from checkpoint, the shape in current model is torch. size mismatch for face_encoder_blocks.0.0.conv_block.0.bias: copying a param with shape torch.Size([8]) from checkpoint, the shape in current model is torch.Size([16]). size mismatch for face_encoder_blocks.0.0.conv_block.1.weight: copying a param with shape torch.Size([8]) from checkpoint, the shape in current model is torch.Size([16] size mismatch for face_encoder_blocks.0.0.conv_block.1.bias: copying a param with shape torch.Size([8]) from checkpoint, the shape in current model is torch.Size([16]). size mismatch for face_encoder_blocks.0.0.conv_block.1.running_mean: copying a param with shape torch.Size([8]) from checkpoint, the shape in current model is torch.Siz size mismatch for face_encoder_blocks.0.0.conv_block.1.running_var: copying a param with shape torch.Size([8]) from checkpoint, the shape in current model is torch.Size size mismatch for face_encoder_blocks.1.0.conv_block.0.weight: copying a param with shape torch.Size([16, 8, 3, 3]) from checkpoint, the shape in current model is torch size mismatch for face_encoder_blocks.1.0.conv_block.0.bias: copying a param with shape torch.Size([16]) from checkpoint, the shape in current model is torch.Size([32]) size mismatch for face_encoder_blocks.1.0.conv_block.1.weight: copying a param with shape torch.Size([16]) from checkpoint, the shape in current model is torch.Size([32 size mismatch for face_encoder_blocks.1.0.conv_block.1.bias: copying a param with shape torch.Size([16]) from checkpoint, the shape in current model is torch.Size([32]) size mismatch for face_encoder_blocks.1.0.conv_block.1.running_mean: copying a param with shape torch.Size([16]) from checkpoint, the shape in current model is torch.Si size mismatch for face_encoder_blocks.1.0.conv_block.1.running_var: copying a param with shape torch.Size([16]) from checkpoint, the shape in current model is torch.Siz size mismatch for face_encoder_blocks.1.1.conv_block.0.weight: copying a param with shape torch.Size([16, 16, 3, 3]) from checkpoint, the shape in current model is torc size mismatch for face_encoder_blocks.1.1.conv_block.0.bias: copying a param with shape torch.Size([16]) from checkpoint, the shape in current model is torch.Size([32]) size mismatch for face_encoder_blocks.1.1.conv_block.1.weight: copying a param with shape torch.Size([16]) from checkpoint, the shape in current model is torch.Size([32 size mismatch for face_encoder_blocks.1.1.conv_block.1.bias: copying a param with shape torch.Size([16]) from checkpoint, the shape in current model is torch.Size([32]) size mismatch for face_encoder_blocks.1.1.conv_block.1.running_mean: copying a param with shape torch.Size([16]) from checkpoint, the shape in current model is torch.Si size mismatch for face_encoder_blocks.1.1.conv_block.1.running_var: copying a param with shape torch.Size([16]) from checkpoint, the shape in current model is torch.Siz size mismatch for face_encoder_blocks.1.2.conv_block.0.weight: copying a param with shape torch.Size([16, 16, 3, 3]) from checkpoint, the shape in current model is torc size mismatch for face_encoder_blocks.1.2.conv_block.0.bias: copying a param with shape torch.Size([16]) from checkpoint, the shape in current model is torch.Size([32]) size mismatch for face_encoder_blocks.1.2.conv_block.1.weight: copying a param with shape torch.Size([16]) from checkpoint, the shape in current model is torch.Size([32 size mismatch for face_encoder_blocks.1.2.conv_block.1.bias: copying a param with shape torch.Size([16]) from checkpoint, the shape in current model is torch.Size([32]) size mismatch for face_encoder_blocks.1.2.conv_block.1.running_mean: copying a param with shape torch.Size([16]) from checkpoint, the shape in current model is torch.Si size mismatch for face_encoder_blocks.1.2.conv_block.1.running_var: copying a param with shape torch.Size([16]) from checkpoint, the shape in current model is torch.Siz size mismatch for face_encoder_blocks.2.0.conv_block.0.weight: copying a param with shape torch.Size([32, 16, 3, 3]) from checkpoint, the shape in current model is torc size mismatch for face_encoder_blocks.2.0.conv_block.0.bias: copying a param with shape torch.Size([32]) from checkpoint, the shape in current model is torch.Size([64]) size mismatch for face_encoder_blocks.2.0.conv_block.1.weight: copying a param with shape torch.Size([32]) from checkpoint, the shape in current model is torch.Size([64 size mismatch for face_encoder_blocks.2.0.conv_block.1.bias: copying a param with shape torch.Size([32]) from checkpoint, the shape in current model is torch.Size([64]) size mismatch for face_encoder_blocks.2.0.conv_block.1.running_mean: copying a param with shape torch.Size([32]) from checkpoint, the shape in current model is torch.Si size mismatch for face_encoder_blocks.2.0.conv_block.1.running_var: copying a param with shape torch.Size([32]) from checkpoint, the shape in current model is torch.Siz size mismatch for face_encoder_blocks.2.1.conv_block.0.weight: copying a param with shape torch.Size([32, 32, 3, 3]) from checkpoint, the shape in current model is torc size mismatch for face_encoder_blocks.2.1.conv_block.0.bias: copying a param with shape torch.Size([32]) from checkpoint, the shape in current model is torch.Size([64]) size mismatch for face_encoder_blocks.2.1.conv_block.1.weight: copying a param with shape torch.Size([32]) from checkpoint, the shape in current model is torch.Size([64 size mismatch for face_encoder_blocks.2.1.conv_block.1.bias: copying a param with shape torch.Size([32]) from checkpoint, the shape in current model is torch.Size([64]) size mismatch for face_encoder_blocks.2.1.conv_block.1.running_mean: copying a param with shape torch.Size([32]) from checkpoint, the shape in current model is torch.Si size mismatch for face_encoder_blocks.2.1.conv_block.1.running_var: copying a param with shape torch.Size([32]) from checkpoint, the shape in current model is torch.Siz size mismatch for face_encoder_blocks.2.2.conv_block.0.weight: copying a param with shape torch.Size([32, 32, 3, 3]) from checkpoint, the shape in current model is torc size mismatch for face_encoder_blocks.2.2.conv_block.0.bias: copying a param with shape torch.Size([32]) from checkpoint, the shape in current model is torch.Size([64]) size mismatch for face_encoder_blocks.2.2.conv_block.1.weight: copying a param with shape torch.Size([32]) from checkpoint, the shape in current model is torch.Size([64 size mismatch for face_encoder_blocks.2.2.conv_block.1.bias: copying a param with shape torch.Size([32]) from checkpoint, the shape in current model is torch.Size([64]) size mismatch for face_encoder_blocks.2.2.conv_block.1.running_mean: copying a param with shape torch.Size([32]) from checkpoint, the shape in current model is torch.Si size mismatch for face_encoder_blocks.2.2.conv_block.1.running_var: copying a param with shape torch.Size([32]) from checkpoint, the shape in current model is torch.Siz size mismatch for face_encoder_blocks.3.0.conv_block.0.weight: copying a param with shape torch.Size([64, 32, 3, 3]) from checkpoint, the shape in current model is torc size mismatch for face_encoder_blocks.3.0.conv_block.0.bias: copying a param with shape torch.Size([64]) from checkpoint, the shape in current model is torch.Size([128] size mismatch for face_encoder_blocks.3.0.conv_block.1.weight: copying a param with shape torch.Size([64]) from checkpoint, the shape in current model is torch.Size([12 size mismatch for face_encoder_blocks.3.0.conv_block.1.bias: copying a param with shape torch.Size([64]) from checkpoint, the shape in current model is torch.Size([128] size mismatch for face_encoder_blocks.3.0.conv_block.1.running_mean: copying a param with shape torch.Size([64]) from checkpoint, the shape in current model is torch.Si size mismatch for face_encoder_blocks.3.0.conv_block.1.running_var: copying a param with shape torch.Size([64]) from checkpoint, the shape in current model is torch.Siz size mismatch for face_encoder_blocks.3.1.conv_block.0.weight: copying a param with shape torch.Size([64, 64, 3, 3]) from checkpoint, the shape in current model is torc size mismatch for face_encoder_blocks.3.1.conv_block.0.bias: copying a param with shape torch.Size([64]) from checkpoint, the shape in current model is torch.Size([128] size mismatch for face_encoder_blocks.3.1.conv_block.1.weight: copying a param with shape torch.Size([64]) from checkpoint, the shape in current model is torch.Size([12 size mismatch for face_encoder_blocks.3.1.conv_block.1.bias: copying a param with shape torch.Size([64]) from checkpoint, the shape in current model is torch.Size([128] size mismatch for face_encoder_blocks.3.1.conv_block.1.running_mean: copying a param with shape torch.Size([64]) from checkpoint, the shape in current model is torch.Si size mismatch for face_encoder_blocks.3.1.conv_block.1.running_var: copying a param with shape torch.Size([64]) from checkpoint, the shape in current model is torch.Siz size mismatch for face_encoder_blocks.3.2.conv_block.0.weight: copying a param with shape torch.Size([64, 64, 3, 3]) from checkpoint, the shape in current model is torc size mismatch for face_encoder_blocks.3.2.conv_block.0.bias: copying a param with shape torch.Size([64]) from checkpoint, the shape in current model is torch.Size([128] size mismatch for face_encoder_blocks.3.2.conv_block.1.weight: copying a param with shape torch.Size([64]) from checkpoint, the shape in current model is torch.Size([12 size mismatch for face_encoder_blocks.3.2.conv_block.1.bias: copying a param with shape torch.Size([64]) from checkpoint, the shape in current model is torch.Size([128] size mismatch for face_encoder_blocks.3.2.conv_block.1.running_mean: copying a param with shape torch.Size([64]) from checkpoint, the shape in current model is torch.Si size mismatch for face_encoder_blocks.3.2.conv_block.1.running_var: copying a param with shape torch.Size([64]) from checkpoint, the shape in current model is torch.Siz size mismatch for face_encoder_blocks.4.0.conv_block.0.weight: copying a param with shape torch.Size([128, 64, 3, 3]) from checkpoint, the shape in current model is tor size mismatch for face_encoder_blocks.4.0.conv_block.0.bias: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([256 size mismatch for face_encoder_blocks.4.0.conv_block.1.weight: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([2 size mismatch for face_encoder_blocks.4.0.conv_block.1.bias: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([256 size mismatch for face_encoder_blocks.4.0.conv_block.1.running_mean: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.S size mismatch for face_encoder_blocks.4.0.conv_block.1.running_var: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Si size mismatch for face_encoder_blocks.4.1.conv_block.0.weight: copying a param with shape torch.Size([128, 128, 3, 3]) from checkpoint, the shape in current model is to size mismatch for face_encoder_blocks.4.1.conv_block.0.bias: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([256 size mismatch for face_encoder_blocks.4.1.conv_block.1.weight: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([2 size mismatch for face_encoder_blocks.4.1.conv_block.1.bias: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([256 size mismatch for face_encoder_blocks.4.1.conv_block.1.running_mean: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.S size mismatch for face_encoder_blocks.4.1.conv_block.1.running_var: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Si size mismatch for face_encoder_blocks.4.2.conv_block.0.weight: copying a param with shape torch.Size([128, 128, 3, 3]) from checkpoint, the shape in current model is to size mismatch for face_encoder_blocks.4.2.conv_block.0.bias: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([256 size mismatch for face_encoder_blocks.4.2.conv_block.1.weight: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([2 size mismatch for face_encoder_blocks.4.2.conv_block.1.bias: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([256 size mismatch for face_encoder_blocks.4.2.conv_block.1.running_mean: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.S size mismatch for face_encoder_blocks.4.2.conv_block.1.running_var: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Si size mismatch for face_encoder_blocks.5.0.conv_block.0.weight: copying a param with shape torch.Size([256, 128, 3, 3]) from checkpoint, the shape in current model is to size mismatch for face_encoder_blocks.5.0.conv_block.0.bias: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([512 size mismatch for face_encoder_blocks.5.0.conv_block.1.weight: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([5 size mismatch for face_encoder_blocks.5.0.conv_block.1.bias: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([512 size mismatch for face_encoder_blocks.5.0.conv_block.1.running_mean: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.S size mismatch for face_encoder_blocks.5.0.conv_block.1.running_var: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Si size mismatch for face_encoder_blocks.5.1.conv_block.0.weight: copying a param with shape torch.Size([256, 256, 3, 3]) from checkpoint, the shape in current model is to size mismatch for face_encoder_blocks.5.1.conv_block.0.bias: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([512 size mismatch for face_encoder_blocks.5.1.conv_block.1.weight: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([5 size mismatch for face_encoder_blocks.5.1.conv_block.1.bias: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([512 size mismatch for face_encoder_blocks.5.1.conv_block.1.running_mean: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.S size mismatch for face_encoder_blocks.5.1.conv_block.1.running_var: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Si size mismatch for face_encoder_blocks.6.0.conv_block.0.weight: copying a param with shape torch.Size([512, 256, 3, 3]) from checkpoint, the shape in current model is to size mismatch for face_encoder_blocks.6.1.conv_block.0.weight: copying a param with shape torch.Size([512, 512, 3, 3]) from checkpoint, the shape in current model is to size mismatch for audio_encoder.11.conv_block.0.weight: copying a param with shape torch.Size([256, 256, 3, 3]) from checkpoint, the shape in current model is torch.Siz size mismatch for audio_encoder.11.conv_block.0.bias: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([512]).
size mismatch for audio_encoder.11.conv_block.1.weight: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([512]). size mismatch for audio_encoder.11.conv_block.1.running_mean: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([51 size mismatch for audio_encoder.11.conv_block.1.running_var: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([512 size mismatch for audio_encoder.12.conv_block.0.weight: copying a param with shape torch.Size([512, 256, 3, 3]) from checkpoint, the shape in current model is torch.Siz size mismatch for face_decoder_blocks.0.0.conv_block.0.weight: copying a param with shape torch.Size([1024, 1024, 1, 1]) from checkpoint, the shape in current model is size mismatch for face_decoder_blocks.0.0.conv_block.0.bias: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch.Size([51 size mismatch for face_decoder_blocks.0.0.conv_block.1.weight: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch.Size([ size mismatch for face_decoder_blocks.0.0.conv_block.1.bias: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch.Size([51 size mismatch for face_decoder_blocks.0.0.conv_block.1.running_mean: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch. size mismatch for face_decoder_blocks.0.0.conv_block.1.running_var: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch.S size mismatch for face_decoder_blocks.1.0.conv_block.0.weight: copying a param with shape torch.Size([2048, 1024, 3, 3]) from checkpoint, the shape in current model is size mismatch for face_decoder_blocks.1.0.conv_block.0.bias: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch.Size([51 size mismatch for face_decoder_blocks.1.0.conv_block.1.weight: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch.Size([ size mismatch for face_decoder_blocks.1.0.conv_block.1.bias: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch.Size([51 size mismatch for face_decoder_blocks.1.0.conv_block.1.running_mean: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch. size mismatch for face_decoder_blocks.1.0.conv_block.1.running_var: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch.S size mismatch for face_decoder_blocks.1.1.conv_block.0.weight: copying a param with shape torch.Size([1024, 1024, 3, 3]) from checkpoint, the shape in current model is size mismatch for face_decoder_blocks.1.1.conv_block.0.bias: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch.Size([51 size mismatch for face_decoder_blocks.1.1.conv_block.1.weight: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch.Size([ size mismatch for face_decoder_blocks.1.1.conv_block.1.bias: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch.Size([51 size mismatch for face_decoder_blocks.1.1.conv_block.1.running_mean: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch. size mismatch for face_decoder_blocks.1.1.conv_block.1.running_var: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch.S size mismatch for face_decoder_blocks.2.0.conv_block.0.weight: copying a param with shape torch.Size([2048, 1024, 3, 3]) from checkpoint, the shape in current model is size mismatch for face_decoder_blocks.2.0.conv_block.0.bias: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch.Size([51 size mismatch for face_decoder_blocks.2.0.conv_block.1.weight: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch.Size([ size mismatch for face_decoder_blocks.2.0.conv_block.1.bias: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch.Size([51 size mismatch for face_decoder_blocks.2.0.conv_block.1.running_mean: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch. size mismatch for face_decoder_blocks.2.0.conv_block.1.running_var: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch.S size mismatch for face_decoder_blocks.2.1.conv_block.0.weight: copying a param with shape torch.Size([1024, 1024, 3, 3]) from checkpoint, the shape in current model is size mismatch for face_decoder_blocks.2.1.conv_block.0.bias: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch.Size([51 size mismatch for face_decoder_blocks.2.1.conv_block.1.weight: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch.Size([ size mismatch for face_decoder_blocks.2.1.conv_block.1.bias: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch.Size([51 size mismatch for face_decoder_blocks.2.1.conv_block.1.running_mean: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch. size mismatch for face_decoder_blocks.2.1.conv_block.1.running_var: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch.S size mismatch for face_decoder_blocks.2.2.conv_block.0.weight: copying a param with shape torch.Size([1024, 1024, 3, 3]) from checkpoint, the shape in current model is size mismatch for face_decoder_blocks.2.2.conv_block.0.bias: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch.Size([51 size mismatch for face_decoder_blocks.2.2.conv_block.1.weight: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch.Size([ size mismatch for face_decoder_blocks.2.2.conv_block.1.bias: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch.Size([51 size mismatch for face_decoder_blocks.2.2.conv_block.1.running_mean: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch. size mismatch for face_decoder_blocks.2.2.conv_block.1.running_var: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch.S size mismatch for face_decoder_blocks.3.0.conv_block.0.weight: copying a param with shape torch.Size([1536, 768, 3, 3]) from checkpoint, the shape in current model is t size mismatch for face_decoder_blocks.3.0.conv_block.0.bias: copying a param with shape torch.Size([768]) from checkpoint, the shape in current model is torch.Size([384 size mismatch for face_decoder_blocks.3.0.conv_block.1.weight: copying a param with shape torch.Size([768]) from checkpoint, the shape in current model is torch.Size([3 size mismatch for face_decoder_blocks.3.0.conv_block.1.bias: copying a param with shape torch.Size([768]) from checkpoint, the shape in current model is torch.Size([384 size mismatch for face_decoder_blocks.3.0.conv_block.1.running_mean: copying a param with shape torch.Size([768]) from checkpoint, the shape in current model is torch.S size mismatch for face_decoder_blocks.3.0.conv_block.1.running_var: copying a param with shape torch.Size([768]) from checkpoint, the shape in current model is torch.Si size mismatch for face_decoder_blocks.3.1.conv_block.0.weight: copying a param with shape torch.Size([768, 768, 3, 3]) from checkpoint, the shape in current model is to size mismatch for face_decoder_blocks.3.1.conv_block.0.bias: copying a param with shape torch.Size([768]) from checkpoint, the shape in current model is torch.Size([384 size mismatch for face_decoder_blocks.3.1.conv_block.1.weight: copying a param with shape torch.Size([768]) from checkpoint, the shape in current model is torch.Size([3 size mismatch for face_decoder_blocks.3.1.conv_block.1.bias: copying a param with shape torch.Size([768]) from checkpoint, the shape in current model is torch.Size([384 size mismatch for face_decoder_blocks.3.1.conv_block.1.running_mean: copying a param with shape torch.Size([768]) from checkpoint, the shape in current model is torch.S size mismatch for face_decoder_blocks.3.1.conv_block.1.running_var: copying a param with shape torch.Size([768]) from checkpoint, the shape in current model is torch.Si size mismatch for face_decoder_blocks.3.2.conv_block.0.weight: copying a param with shape torch.Size([768, 768, 3, 3]) from checkpoint, the shape in current model is to size mismatch for face_decoder_blocks.3.2.conv_block.0.bias: copying a param with shape torch.Size([768]) from checkpoint, the shape in current model is torch.Size([384 size mismatch for face_decoder_blocks.3.2.conv_block.1.weight: copying a param with shape torch.Size([768]) from checkpoint, the shape in current model is torch.Size([3 size mismatch for face_decoder_blocks.3.2.conv_block.1.bias: copying a param with shape torch.Size([768]) from checkpoint, the shape in current model is torch.Size([384 size mismatch for face_decoder_blocks.3.2.conv_block.1.running_mean: copying a param with shape torch.Size([768]) from checkpoint, the shape in current model is torch.S size mismatch for face_decoder_blocks.3.2.conv_block.1.running_var: copying a param with shape torch.Size([768]) from checkpoint, the shape in current model is torch.Si size mismatch for face_decoder_blocks.4.0.conv_block.0.weight: copying a param with shape torch.Size([1024, 512, 3, 3]) from checkpoint, the shape in current model is t size mismatch for face_decoder_blocks.4.0.conv_block.0.bias: copying a param with shape torch.Size([512]) from checkpoint, the shape in current model is torch.Size([256 size mismatch for face_decoder_blocks.4.0.conv_block.1.weight: copying a param with shape torch.Size([512]) from checkpoint, the shape in current model is torch.Size([2 size mismatch for face_decoder_blocks.4.0.conv_block.1.bias: copying a param with shape torch.Size([512]) from checkpoint, the shape in current model is torch.Size([256 size mismatch for face_decoder_blocks.4.0.conv_block.1.running_mean: copying a param with shape torch.Size([512]) from checkpoint, the shape in current model is torch.S size mismatch for face_decoder_blocks.4.0.conv_block.1.running_var: copying a param with shape torch.Size([512]) from checkpoint, the shape in current model is torch.Si size mismatch for face_decoder_blocks.4.1.conv_block.0.weight: copying a param with shape torch.Size([512, 512, 3, 3]) from checkpoint, the shape in current model is to size mismatch for face_decoder_blocks.4.1.conv_block.0.bias: copying a param with shape torch.Size([512]) from checkpoint, the shape in current model is torch.Size([256 size mismatch for face_decoder_blocks.4.1.conv_block.1.weight: copying a param with shape torch.Size([512]) from checkpoint, the shape in current model is torch.Size([2 size mismatch for face_decoder_blocks.4.1.conv_block.1.bias: copying a param with shape torch.Size([512]) from checkpoint, the shape in current model is torch.Size([256 size mismatch for face_decoder_blocks.4.1.conv_block.1.running_mean: copying a param with shape torch.Size([512]) from checkpoint, the shape in current model is torch.S size mismatch for face_decoder_blocks.4.1.conv_block.1.running_var: copying a param with shape torch.Size([512]) from checkpoint, the shape in current model is torch.Si size mismatch for face_decoder_blocks.4.2.conv_block.0.weight: copying a param with shape torch.Size([512, 512, 3, 3]) from checkpoint, the shape in current model is to size mismatch for face_decoder_blocks.4.2.conv_block.0.bias: copying a param with shape torch.Size([512]) from checkpoint, the shape in current model is torch.Size([256 size mismatch for face_decoder_blocks.4.2.conv_block.1.weight: copying a param with shape torch.Size([512]) from checkpoint, the shape in current model is torch.Size([2 size mismatch for face_decoder_blocks.4.2.conv_block.1.bias: copying a param with shape torch.Size([512]) from checkpoint, the shape in current model is torch.Size([256 size mismatch for face_decoder_blocks.4.2.conv_block.1.running_mean: copying a param with shape torch.Size([512]) from checkpoint, the shape in current model is torch.S size mismatch for face_decoder_blocks.4.2.conv_block.1.running_var: copying a param with shape torch.Size([512]) from checkpoint, the shape in current model is torch.Si size mismatch for face_decoder_blocks.5.0.conv_block.0.weight: copying a param with shape torch.Size([640, 256, 3, 3]) from checkpoint, the shape in current model is to size mismatch for face_decoder_blocks.5.0.conv_block.0.bias: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([128 size mismatch for face_decoder_blocks.5.0.conv_block.1.weight: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([1 size mismatch for face_decoder_blocks.5.0.conv_block.1.bias: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([128 size mismatch for face_decoder_blocks.5.0.conv_block.1.running_mean: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.S size mismatch for face_decoder_blocks.5.0.conv_block.1.running_var: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Si size mismatch for face_decoder_blocks.5.1.conv_block.0.weight: copying a param with shape torch.Size([256, 256, 3, 3]) from checkpoint, the shape in current model is to size mismatch for face_decoder_blocks.5.1.conv_block.0.bias: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([128 size mismatch for face_decoder_blocks.5.1.conv_block.1.weight: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([1 size mismatch for face_decoder_blocks.5.1.conv_block.1.bias: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([128 size mismatch for face_decoder_blocks.5.1.conv_block.1.running_mean: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.S size mismatch for face_decoder_blocks.5.1.conv_block.1.running_var: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Si size mismatch for face_decoder_blocks.5.2.conv_block.0.weight: copying a param with shape torch.Size([256, 256, 3, 3]) from checkpoint, the shape in current model is to size mismatch for face_decoder_blocks.5.2.conv_block.0.bias: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([128 size mismatch for face_decoder_blocks.5.2.conv_block.1.weight: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([1 size mismatch for face_decoder_blocks.5.2.conv_block.1.bias: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([128 size mismatch for face_decoder_blocks.5.2.conv_block.1.running_mean: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.S size mismatch for face_decoder_blocks.5.2.conv_block.1.running_var: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Si size mismatch for face_decoder_blocks.6.0.conv_block.0.weight: copying a param with shape torch.Size([320, 128, 3, 3]) from checkpoint, the shape in current model is to size mismatch for face_decoder_blocks.6.0.conv_block.0.bias: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([64] size mismatch for face_decoder_blocks.6.0.conv_block.1.weight: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([6 size mismatch for face_decoder_blocks.6.0.conv_block.1.bias: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([64] size mismatch for face_decoder_blocks.6.0.conv_block.1.running_mean: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.S size mismatch for face_decoder_blocks.6.0.conv_block.1.running_var: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Si size mismatch for face_decoder_blocks.6.1.conv_block.0.weight: copying a param with shape torch.Size([128, 128, 3, 3]) from checkpoint, the shape in current model is to size mismatch for face_decoder_blocks.6.1.conv_block.0.bias: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([64] size mismatch for face_decoder_blocks.6.1.conv_block.1.weight: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([6 size mismatch for face_decoder_blocks.6.1.conv_block.1.bias: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([64] size mismatch for face_decoder_blocks.6.1.conv_block.1.running_mean: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.S size mismatch for face_decoder_blocks.6.1.conv_block.1.running_var: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Si size mismatch for face_decoder_blocks.6.2.conv_block.0.weight: copying a param with shape torch.Size([128, 128, 3, 3]) from checkpoint, the shape in current model is to size mismatch for face_decoder_blocks.6.2.conv_block.0.bias: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([64] size mismatch for face_decoder_blocks.6.2.conv_block.1.weight: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([6 size mismatch for face_decoder_blocks.6.2.conv_block.1.bias: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([64] size mismatch for face_decoder_blocks.6.2.conv_block.1.running_mean: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.S size mismatch for face_decoder_blocks.6.2.conv_block.1.running_var: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Si size mismatch for output_block.0.conv_block.0.weight: copying a param with shape torch.Size([16, 40, 3, 3]) from checkpoint, the shape in current model is torch.Size([3 size mismatch for output_block.0.conv_block.0.bias: copying a param with shape torch.Size([16]) from checkpoint, the shape in current model is torch.Size([32]).
size mismatch for output_block.0.conv_block.1.weight: copying a param with shape torch.Size([16]) from checkpoint, the shape in current model is torch.Size([32]).
size mismatch for output_block.0.conv_block.1.bias: copying a param with shape torch.Size([16]) from checkpoint, the shape in current model is torch.Size([32]).
size mismatch for output_block.0.conv_block.1.running_mean: copying a param with shape torch.Size([16]) from checkpoint, the shape in current model is torch.Size([32]). size mismatch for output_block.0.conv_block.1.running_var: copying a param with shape torch.Size([16]) from checkpoint, the shape in current model is torch.Size([32]).< size mismatch for output_block.1.weight: copying a param with shape torch.Size([3, 16, 1, 1]) from checkpoint, the shape in current model is torch.Size([3, 32, ffmpeg version 4.4.2-0ubuntu0.22.04.1 Copyright (c) 2000-2021 the FFmpeg developers "face_encoder_blocks.2.3.conv_block.1.bias", "face_encoder_blocks.2.3.conv_block.1.running_mean", "face_encoder_blocks.2.3.conv_block.1.running_var".
ock.1.bias", "audio_refine.0.conv_block.1.running_mean", "audio_refine.0.conv_block.1.running_var", "audio_refine.0.conv_block.1.num_batches_tracked", "audio_refine.1.conv_bloc k.0.weight", "audio_refine.1.conv_block.0.bias", "audio_refine.1.conv_block.1.weight", "audio_refine.1.conv_block.1.bias", "audio_refine.1.conv_block.1.running_mean", "audio_re fine.1.conv_block.1.running_var", "audio_refine.1.conv_block.1.num_batches_tracked", "face_encoder_blocks.7.0.conv_block.0.weight", "face_encoder_blocks.7.0.conv_block.0.bias", "face_encoder_blocks.7.0.conv_block.1.weight", "face_encoder_blocks.7.0.conv_block.1.bias", "face_encoder_blocks.7.0.conv_block.1.running_mean", "face_encoder_blocks.7.0.conv_ block.1.running_var", "face_encoder_blocks.7.0.conv_block.1.num_batches_tracked", "face_encoder_blocks.7.1.conv_block.0.weight", "face_encoder_blocks.7.1.conv_block.0.bias", "f ace_encoder_blocks.7.1.conv_block.1.weight", "face_encoder_blocks.7.1.conv_block.1.bias", "face_encoder_blocks.7.1.conv_block.1.running_mean", "face_encoder_blocks.7.1.conv_blo ck.1.running_var", "face_encoder_blocks.7.1.conv_block.1.num_batches_tracked", "face_encoder_blocks.7.2.conv_block.0.weight", "face_encoder_blocks.7.2.conv_block.0.bias", "face encoder_blocks.7.2.conv_block.1.weight", "face_encoder_blocks.7.2.conv_block.1.bias", "face_encoder_blocks.7.2.conv_block.1.running_mean", "face_encoder_blocks.7.2.conv_block. 1.running_var", "face_encoder_blocks.7.2.conv_block.1.num_batches_tracked", "face_encoder_blocks.8.0.conv_block.0.weight", "face_encoder_blocks.8.0.conv_block.0.bias", "face_en coder_blocks.8.0.conv_block.1.weight", "face_encoder_blocks.8.0.conv_block.1.bias", "face_encoder_blocks.8.0.conv_block.1.running_mean", "face_encoder_blocks.8.0.conv_block.1.r unning_var", "face_encoder_blocks.8.0.conv_block.1.num_batches_tracked", "face_encoder_blocks.8.1.conv_block.0.weight", "face_encoder_blocks.8.1.conv_block.0.bias", "face_encod er_blocks.8.1.conv_block.1.weight", "face_encoder_blocks.8.1.conv_block.1.bias", "face_encoder_blocks.8.1.conv_block.1.running_mean", "face_encoder_blocks.8.1.conv_block.1.runn ing_var", "face_encoder_blocks.8.1.conv_block.1.num_batches_tracked", "face_encoder_blocks.3.3.conv_block.0.weight", "face_encoder_blocks.3.3.conv_block.0.bias", "face_encoder blocks.3.3.conv_block.1.weight", "face_encoder_blocks.3.3.conv_block.1.bias", "face_encoder_blocks.3.3.conv_block.1.running_mean", "face_encoder_blocks.3.3.conv_block.1.running var", "face_encoder_blocks.3.3.conv_block.1.num_batches_tracked", "face_encoder_blocks.5.2.conv_block.0.weight", "face_encoder_blocks.5.2.conv_block.0.bias", "face_encoder_blo cks.5.2.conv_block.1.weight", "face_encoder_blocks.5.2.conv_block.1.bias", "face_encoder_blocks.5.2.conv_block.1.running_mean", "face_encoder_blocks.5.2.conv_block.1.running_va r", "face_encoder_blocks.5.2.conv_block.1.num_batches_tracked", "face_encoder_blocks.6.2.conv_block.0.weight", "face_encoder_blocks.6.2.conv_block.0.bias", "face_encoder_blocks .6.2.conv_block.1.weight", "face_encoder_blocks.6.2.conv_block.1.bias", "face_encoder_blocks.6.2.conv_block.1.running_mean", "face_encoder_blocks.6.2.conv_block.1.running_var", "face_encoder_blocks.6.2.conv_block.1.num_batches_tracked", "audio_encoder.13.conv_block.0.weight", "audio_encoder.13.conv_block.0.bias", "audio_encoder.13.conv_block.1.weight ", "audio_encoder.13.conv_block.1.bias", "audio_encoder.13.conv_block.1.running_mean", "audio_encoder.13.conv_block.1.running_var", "audio_encoder.13.conv_block.1.num_batches_t racked", "audio_encoder.14.conv_block.0.weight", "audio_encoder.14.conv_block.0.bias", "audio_encoder.14.conv_block.1.weight", "audio_encoder.14.conv_block.1.bias", "audio_enco der.14.conv_block.1.running_mean", "audio_encoder.14.conv_block.1.running_var", "audio_encoder.14.conv_block.1.num_batches_tracked", "audio_encoder.15.conv_block.0.weight", "au dio_encoder.15.conv_block.0.bias", "audio_encoder.15.conv_block.1.weight", "audio_encoder.15.conv_block.1.bias", "audio_encoder.15.conv_block.1.running_mean", "audio_encoder.15 .conv_block.1.running_var", "audio_encoder.15.conv_block.1.num_batches_tracked", "audio_encoder.16.conv_block.0.weight", "audio_encoder.16.conv_block.0.bias", "audio_encoder.16 .conv_block.1.weight", "audio_encoder.16.conv_block.1.bias", "audio_encoder.16.conv_block.1.running_mean", "audio_encoder.16.conv_block.1.running_var", "audio_encoder.16.conv_b lock.1.num_batches_tracked", "face_decoder_blocks.7.0.conv_block.0.weight", "face_decoder_blocks.7.0.conv_block.0.bias", "face_decoder_blocks.7.0.conv_block.1.weight", "face_de coder_blocks.7.0.conv_block.1.bias", "face_decoder_blocks.7.0.conv_block.1.running_mean", "face_decoder_blocks.7.0.conv_block.1.running_var", "face_decoder_blocks.7.0.conv_bloc k.1.num_batches_tracked", "face_decoder_blocks.7.1.conv_block.0.weight", "face_decoder_blocks.7.1.conv_block.0.bias", "face_decoder_blocks.7.1.conv_block.1.weight", "face_decod er_blocks.7.1.conv_block.1.bias", "face_decoder_blocks.7.1.conv_block.1.running_mean", "face_decoder_blocks.7.1.conv_block.1.running_var", "face_decoder_blocks.7.1.conv_block.1 .num_batches_tracked", "face_decoder_blocks.7.2.conv_block.0.weight", "face_decoder_blocks.7.2.conv_block.0.bias", "face_decoder_blocks.7.2.conv_block.1.weight", "face_decoder blocks.7.2.conv_block.1.bias", "face_decoder_blocks.7.2.conv_block.1.running_mean", "face_decoder_blocks.7.2.conv_block.1.running_var", "face_decoder_blocks.7.2.conv_block.1.nu m_batches_tracked", "face_decoder_blocks.8.0.conv_block.0.weight", "face_decoder_blocks.8.0.conv_block.0.bias", "face_decoder_blocks.8.0.conv_block.1.weight", "face_decoder_blo cks.8.0.conv_block.1.bias", "face_decoder_blocks.8.0.conv_block.1.running_mean", "face_decoder_blocks.8.0.conv_block.1.running_var", "face_decoder_blocks.8.0.conv_block.1.num_b atches_tracked", "face_decoder_blocks.8.1.conv_block.0.weight", "face_decoder_blocks.8.1.conv_block.0.bias", "face_decoder_blocks.8.1.conv_block.1.weight", "face_decoder_blocks .8.1.conv_block.1.bias", "face_decoder_blocks.8.1.conv_block.1.running_mean", "face_decoder_blocks.8.1.conv_block.1.running_var", "face_decoder_blocks.8.1.conv_block.1.num_batc hes_tracked", "face_decoder_blocks.8.2.conv_block.0.weight", "face_decoder_blocks.8.2.conv_block.0.bias", "face_decoder_blocks.8.2.conv_block.1.weight", "face_decoder_blocks.8. 2.conv_block.1.bias", "face_decoder_blocks.8.2.conv_block.1.running_mean", "face_decoder_blocks.8.2.conv_block.1.running_var", "face_decoder_blocks.8.2.conv_block.1.num_batches _tracked".
Size([16, 6, 7, 7]).

).

e([16]).
([16]).
.Size([32, 16, 3, 3]).
.
]).
.
ze([32]).
e([32]).
h.Size([32, 32, 3, 3]).
.
]).
.
ze([32]).
e([32]).
h.Size([32, 32, 3, 3]).
.
]).
.
ze([32]).
e([32]).
h.Size([64, 32, 3, 3]).
.
]).
.
ze([64]).
e([64]).
h.Size([64, 64, 3, 3]).
.
]).
.
ze([64]).
e([64]).
h.Size([64, 64, 3, 3]).
.
]).
.
ze([64]).
e([64]).
h.Size([128, 64, 3, 3]).
).
8]).
).
ze([128]).
e([128]).
h.Size([128, 128, 3, 3]).
).
8]).
).
ze([128]).
e([128]).
h.Size([128, 128, 3, 3]).
).
8]).
).
ze([128]).
e([128]).
ch.Size([256, 128, 3, 3]).
]).
56]).
]).
ize([256]).
ze([256]).
rch.Size([256, 256, 3, 3]).
]).
56]).
]).
ize([256]).
ze([256]).
rch.Size([256, 256, 3, 3]).
]).
56]).
]).
ize([256]).
ze([256]).
rch.Size([512, 256, 3, 3]).
]).
12]).
]).
ize([512]).
ze([512]).
rch.Size([512, 512, 3, 3]).
]).
12]).
]).
ize([512]).
ze([512]).
rch.Size([512, 512, 3, 3]).
rch.Size([512, 512, 1, 1]).
e([512, 256, 3, 3]).
r> 2]).
]).
e([512, 512, 1, 1]).
torch.Size([512, 512, 1, 1]).
2]).
512]).
2]).
Size([512]).
ize([512]).
torch.Size([1024, 512, 3, 3]).
2]).
512]).
2]).
Size([512]).
ize([512]).
torch.Size([512, 512, 3, 3]).
2]).
512]).
2]).
Size([512]).
ize([512]).
torch.Size([1024, 512, 3, 3]).
2]).
512]).
2]).
Size([512]).
ize([512]).
torch.Size([512, 512, 3, 3]).
2]).
512]).
2]).
Size([512]).
ize([512]).
torch.Size([512, 512, 3, 3]).
2]).
512]).
2]).
Size([512]).
ize([512]).
orch.Size([768, 384, 3, 3]).
]).
84]).
]).
ize([384]).
ze([384]).
rch.Size([384, 384, 3, 3]).
]).
84]).
]).
ize([384]).
ze([384]).
rch.Size([384, 384, 3, 3]).
]).
84]).
]).
ize([384]).
ze([384]).
orch.Size([512, 256, 3, 3]).
]).
56]).
]).
ize([256]).
ze([256]).
rch.Size([256, 256, 3, 3]).
]).
56]).
]).
ize([256]).
ze([256]).
rch.Size([256, 256, 3, 3]).
]).
56]).
]).
ize([256]).
ze([256]).
rch.Size([320, 128, 3, 3]).
]).
28]).
]).
ize([128]).
ze([128]).
rch.Size([128, 128, 3, 3]).
]).
28]).
]).
ize([128]).
ze([128]).
rch.Size([128, 128, 3, 3]).
]).
28]).
]).
ize([128]).
ze([128]).
rch.Size([160, 64, 3, 3]).
).
4]).
).
ize([64]).
ze([64]).
rch.Size([64, 64, 3, 3]).
).
4]).
).
ize([64]).
ze([64]).
rch.Size([64, 64, 3, 3]).
).
4]).
).
ize([64]).
ze([64]).
2, 80, 3, 3]).

br> 1, 1]).

@H-Liu1997

@mnbv7758 Thanks for your interest!

if you want to try high-quality mode, please consider:

use the saved HD video in line 369 in app.py

replace line 370 - line 374 with a HD lipsync model, in our HD demo we use in-lab model. for open-source version, I suggest you take a look on wav2lip-hd

This is line 369: video_temp_path = os.path.join(save_dir, f"audio_{idx}retri{counter}.mp4") of app.py. what part to change.
For number 2, can you suggest example lines of code to replace line 370-374

	if trans_similarity < threshold_trans:
	if np.sum(position_similarity < threshold_position) >= 45 and np.sum(velocity_similarity < threshold_velocity) >= 45:
	graph.add_edge(i, j, is_continue=0)