wav2lip_288x288 introduction

This is a project about talking faces. We use288X288 sized facial images for training, which can generate720p, 1080p, 2k Digital Humanhuman videos. We have done the following work:

1. Add video cutting codes.
2. Add filelists to generate code.
3. Trained 600 people, 30 hours, and over 30000 pieces of data.
4. Open sourced the checkpoint for a discriminator with 150000 steps and a val_rass of 0.28.
5. Open sourced a checkpoint for a generator with 360000 steps and a val_rass of 0.25.
6. Dear friends, you can load pre training weights for easy subsequent training.

wav2lip-288x288 Project situation

Video | Project Page | Code

checkpoints for wav2lip_288x288 https://pan.baidu.com/s/1ks53RXFzN56Ksjpxspiwyw?pwd=lzzx

The following pictures are comparison images of the training generator training 300000 steps.

The following images show the loss values of training the discriminator for 300000 steps.

The following images show the loss values of training the generator for 300000 steps.

Release Plan

For the wav2lip series, we will continue to train and release higher definition weights in the future. The plan is as follows: Pre training checkpoints for wav2lip_288x288 will be released in January 2025. Pre training checkpoints for wav2lip_384x384 will be released in February 2025. Pre training checkpoints for wav2lip_576x576 or 512x512 will be released in June 2025.

Citing

Thank the two authors, Thank you for their wonderful work. https://github.com/primepake/wav2lip_288x288 https://github.com/Rudrabha/Wav2Lip

Disclaimers

This repositories made by langzizhixin from Langzizhixin Technology company 2025.1.1 , in Chengdu, China . The above code and weights can only be used for personal/research/non-commercial purposes. If you need a higher definition model, please contact me by email [email protected] , or add WeChat for communication: langzizhixinkeji