# ZiqiaoPeng/SyncTalk

[CVPR 2024] This is the official source for our paper "SyncTalk: The Devil is in the Synchronization for Talking Head Synthesis"

Repository: https://github.com/ZiqiaoPeng/SyncTalk
Canonical: https://ross.abutalabs.com/products/synctalk
Homepage: https://ziqiaopeng.github.io/synctalk/
Language: Python
License: NOASSERTION
License Family: other
Topics: talking-face-generation, talking-head, audio-driven-talking-face, talking-face, cvpr, cvpr2024
Last push: 2025-09-18T15:00:06+00:00

## Health v2 (maintenance only)
Score: 46/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 42, release rhythm 35, longevity 72
- inputs: {"age_days": 1008, "days_push": 349, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1626, forks 196 (observed 2026-08-28T04:05:12.994281+00:00)

## What it is
SyncTalk is the official PyTorch implementation of a CVPR 2024 paper that synthesizes speech-driven, synchronized talking head videos using NeRF-based tri-plane hash representations. It generates lip-synced speech, facial expressions, stable head poses, and high-resolution hair and torso details from a portrait image and audio input.

## Use cases
- generate a talking head video from a photo and audio clip
- synthesize lip-synced avatar videos from speech
- research code for audio-driven talking face generation
- create high-fidelity digital human videos
- reproduce CVPR 2024 talking head synthesis results
- animate a portrait with realistic head poses and expressions

## When to choose
- you need research-grade, synchronized talking head synthesis with identity preservation
- you want NeRF-based 3D-aware portrait animation with expressions and head pose
- you are reproducing or building on the SyncTalk paper

## When to avoid
- you need a production-ready, easy-to-deploy avatar API rather than research code
- you lack a CUDA GPU or cannot handle a heavy PyTorch/CUDA setup
- you need real-time 2D generation - the author's SyncTalk_2D may fit better

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning, video-processing, audio-processing, speech-recognition, computer-vision
- domain: computer-vision, deep-learning, artificial-intelligence
- platform: windows, python
- tags: talking-head-synthesis, nerf, audio-driven, face-generation, cvpr-2024, research-code, neural-rendering, audio, video, linux, gpu

## Member repositories
- ZiqiaoPeng/SyncTalk (main) score 46

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:12.994281+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T03:48:26.828712+00:00, confidence not recorded.
  - readme: https://github.com/ZiqiaoPeng/SyncTalk (fetched 2026-08-28T04:05:12.994281+00:00, sha 2161ac32a594)
  - homepage: https://ziqiaopeng.github.io/synctalk/ (fetched 2026-08-29T11:21:15.680821+00:00, sha fb3a61e0b341)
- Data as of 2026-08-30T08:39:29.467469+00:00.