ComfyUI Extension: LatentSync

Authored by jiafuzeng

Created

Updated

0 stars

We present LatentSync, an end-to-end lip-sync method based on audio-conditioned latent diffusion models without any intermediate motion representation, diverging from previous diffusion-based lip-sync methods based on pixel-space diffusion or two-stage generation. Our framework can leverage the powerful capabilities of Stable Diffusion to directly model complex audio-visual correlations.

Custom Nodes (0)

    README

    LatentSync

    <!-- <div align="center"> -->

    arXiv arXiv arXiv <a href="https://replicate.com/lucataco/latentsync"><img src="https://replicate.com/lucataco/latentsync/badge" alt="Replicate"></a>

    <!-- </div> -->

    🔥 Updates

    • 2025/03/14: We released LatentSync 1.5, which (1) improves temporal consistency via adding temporal layer, (2) improves performance on Chinese videos and (3) reduces the VRAM requirement of the stage2 training to 23 GB through a series of optimizations. Learn more details at this.

    📖 Introduction

    We present LatentSync, an end-to-end lip-sync method based on audio-conditioned latent diffusion models without any intermediate motion representation, diverging from previous diffusion-based lip-sync methods based on pixel-space diffusion or two-stage generation. Our framework can leverage the powerful capabilities of Stable Diffusion to directly model complex audio-visual correlations.

    🏗️ Framework

    <p align="center"> <img src="docs/framework.png" width=100%> <p>

    LatentSync uses the Whisper to convert melspectrogram into audio embeddings, which are then integrated into the U-Net via cross-attention layers. The reference and masked frames are channel-wise concatenated with noised latents as the input of U-Net. In the training process, we use a one-step method to get estimated clean latents from predicted noises, which are then decoded to obtain the estimated clean frames. The TREPA, LPIPS and SyncNet losses are added in the pixel space.

    🎬 Demo

    <table class="center"> <tr style="font-weight: bolder;text-align:center;"> <td width="50%"><b>Original video</b></td> <td width="50%"><b>Lip-synced video</b></td> </tr> <tr> <td> <video src=https://github.com/user-attachments/assets/ff3a84da-dc9b-498a-950f-5c54f58dd5c5 controls preload></video> </td> <td> <video src=https://github.com/user-attachments/assets/150e00fd-381e-4421-a478-a9ea3d1212a8 controls preload></video> </td> </tr> <tr> <td> <video src=https://github.com/user-attachments/assets/32c830a9-4d7d-4044-9b33-b184d8e11010 controls preload></video> </td> <td> <video src=https://github.com/user-attachments/assets/84e4fe9d-b108-44a4-8712-13a012348145 controls preload></video> </td> </tr> <tr> <td> <video src=https://github.com/user-attachments/assets/7510a448-255a-44ee-b093-a1b98bd3961d controls preload></video> </td> <td> <video src=https://github.com/user-attachments/assets/6150c453-c559-4ae0-bb00-c565f135ff41 controls preload></video> </td> </tr> <tr> <td width=300px> <video src=https://github.com/user-attachments/assets/0f7f9845-68b2-4165-bd08-c7bbe01a0e52 controls preload></video> </td> <td width=300px> <video src=https://github.com/user-attachments/assets/c34fe89d-0c09-4de3-8601-3d01229a69e3 controls preload></video> </td> </tr> <tr> <td> <video src=https://github.com/user-attachments/assets/7ce04d50-d39f-4154-932a-ec3a590a8f64 controls preload></video> </td> <td> <video src=https://github.com/user-attachments/assets/70bde520-42fa-4a0e-b66c-d3040ae5e065 controls preload></video> </td> </tr> </table>

    (Photorealistic videos are filmed by contracted models, and anime videos are from VASA-1 and EMO)

    📑 Open-source Plan

    • [x] Inference code and checkpoints
    • [x] Data processing pipeline
    • [x] Training code

    🔧 Setting up the Environment

    Install the required packages and download the checkpoints via:

    source setup_env.sh
    

    If the download is successful, the checkpoints should appear as follows:

    ./checkpoints/
    |-- latentsync_unet.pt
    |-- latentsync_syncnet.pt
    |-- whisper
    |   `-- tiny.pt
    |-- auxiliary
    |   |-- 2DFAN4-cd938726ad.zip
    |   |-- i3d_torchscript.pt
    |   |-- koniq_pretrained.pkl
    |   |-- s3fd-619a316812.pth
    |   |-- sfd_face.pth
    |   |-- syncnet_v2.model
    |   |-- vgg16-397923af.pth
    |   `-- vit_g_hybrid_pt_1200e_ssv2_ft.pth
    

    These already include all the checkpoints required for latentsync training and inference. If you just want to try inference, you only need to download latentsync_unet.pt and tiny.pt from our HuggingFace repo

    🚀 Inference

    There are two ways to perform inference, and both require 6.5 GB of VRAM.

    1. Gradio App

    Run the Gradio app for inference:

    python gradio_app.py
    

    2. Command Line Interface

    Run the script for inference:

    ./inference.sh
    

    You can change the parameters inference_steps and guidance_scale to see more results.

    🔄 Data Processing Pipeline

    The complete data processing pipeline includes the following steps:

    1. Remove the broken video files.
    2. Resample the video FPS to 25, and resample the audio to 16000 Hz.
    3. Scene detect via PySceneDetect.
    4. Split each video into 5-10 second segments.
    5. Remove videos where the face is smaller than 256 $\times$ 256, as well as videos with more than one face.
    6. Affine transform the faces according to the landmarks detected by face-alignment, then resize to 256 $\times$ 256.
    7. Remove videos with sync confidence score lower than 3, and adjust the audio-visual offset to 0.
    8. Calculate hyperIQA score, and remove videos with scores lower than 40.

    Run the script to execute the data processing pipeline:

    ./data_processing_pipeline.sh
    

    You can change the parameter input_dir in the script to specify the data directory to be processed. The processed data will be saved in the high_visual_quality directory. Each step will generate a new directory to prevent the need to redo the entire pipeline in case the process is interrupted by an unexpected error.

    🏋️‍♂️ Training U-Net

    Before training, you must process the data as described above and download all the checkpoints. We released a pretrained SyncNet with 94% accuracy on the VoxCeleb2 dataset for the supervision of U-Net training. Note that this SyncNet is trained on affine transformed videos, so when using or evaluating this SyncNet, you need to perform affine transformation on the video first (the code of affine transformation is included in the data processing pipeline).

    If all the preparations are complete, you can train the U-Net with the following script:

    ./train_unet.sh
    

    You should change the parameters in U-Net config file to specify the data directory, checkpoint save path, and other training hyperparameters.

    🏋️‍♂️ Training SyncNet

    In case you want to train SyncNet on your own datasets, you can run the following script. The data processing pipeline for SyncNet is the same as U-Net.

    ./train_syncnet.sh
    

    After validations_steps training, the loss charts will be saved in train_output_dir. They contain both the training and validation loss.

    📊 Evaluation

    You can evaluate the sync confidence score of a generated video by running the following script:

    ./eval/eval_sync_conf.sh
    

    You can evaluate the accuracy of SyncNet on a dataset by running the following script:

    ./eval/eval_syncnet_acc.sh
    

    🙏 Acknowledgement

    Thanks for their generous contributions to the open-source community.