# bytedance/LatentSync

Taming Stable Diffusion for Lip Sync!

Repository: https://github.com/bytedance/LatentSync
Canonical: https://ross.abutalabs.com/products/latentsync
Homepage: https://arxiv.org/abs/2412.09262
Language: Python
License: Apache-2.0
License Family: permissive
Topics: research, diffusion-models, lipsync, video-gen, virtual-avatars
Last push: 2025-06-20T07:36:58+00:00

## Health v2 (maintenance only)
Score: 33/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 27, release rhythm 35, longevity 45
- inputs: {"age_days": 630, "days_push": 439, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 6026, forks 977 (observed 2026-08-28T04:09:34.348298+00:00)

## What it is
LatentSync is an end-to-end lip-sync framework from ByteDance based on audio-conditioned latent diffusion models, using Stable Diffusion to directly model audio-visual correlations without intermediate motion representations. It uses Whisper audio embeddings, cross-attention integration into U-Net, and SyncNet supervision (StableSyncNet) for accurate lip synchronization in videos.

## Use cases
- sync lip movements in a video to new audio
- dub a video into another language with matching lips
- create talking avatar videos from a portrait
- generate lip-synced talking head videos
- retime a person's mouth to match a voiceover
- animate a face in a video to match speech audio

## When to choose
- you need high-quality, research-grade lip sync on videos
- you want an open-source alternative to commercial lip-sync tools
- you have a GPU and want local, self-hosted inference
- you need multilingual lip sync including Chinese

## When to avoid
- you lack a GPU or VRAM headroom for diffusion inference
- you need real-time lip sync on live streams
- you only need simple face swaps rather than audio-driven sync
- you need a production API with managed scaling

## Facets
- artifact type: library
- maturity: active
- function: video-processing, machine-learning, deep-learning, speech-recognition, image-processing
- domain: deep-learning, computer-vision, artificial-intelligence
- platform: python
- tags: lip-sync, latent-diffusion, stable-diffusion, talking-head, audio-driven, virtual-avatars, video-generation, syncnet, whisper, video, audio, linux, gpu

## Member repositories
- bytedance/LatentSync (main) score 33

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:09:34.348298+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 2026-08-29T17:49:14.523135+00:00, confidence not recorded.
  - readme: https://github.com/bytedance/LatentSync (fetched 2026-08-28T04:09:34.348298+00:00, sha 5e064aacc127)
  - homepage: https://arxiv.org/abs/2412.09262 (fetched 2026-08-29T08:45:32.285280+00:00, sha ba2c3f1c310f)
  - site_page: https://info.arxiv.org/about/donate.html (fetched 2026-08-29T08:45:32.294640+00:00, sha cca9c3a11c56)
  - site_page: https://info.arxiv.org/about/ourmembers.html (fetched 2026-08-29T08:45:32.298227+00:00, sha 47cbc55ff1de)
  - site_page: https://info.arxiv.org/about (fetched 2026-08-29T08:45:32.300723+00:00, sha a1f16f915a9a)
  - site_page: https://info.arxiv.org/labs/index.html (fetched 2026-08-29T08:45:32.296486+00:00, sha b14a8d05a0ec)
- Data as of 2026-08-30T08:39:29.467469+00:00.