# ali-vilab/dreamtalk

Official implementations for paper: DreamTalk: When Expressive Talking Head Generation Meets Diffusion Probabilistic Models

Repository: https://github.com/ali-vilab/dreamtalk
Canonical: https://ross.abutalabs.com/products/dreamtalk
Homepage: https://dreamtalk-project.github.io/
Language: Python
License: MIT
License Family: permissive
Topics: audio-visual-learning, face-animation, talking-head, video-generation
Last push: 2024-01-15T19:11:17+00:00

## Health v2 (maintenance only)
Score: 26/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 69
- inputs: {"age_days": 979, "days_push": 961, "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 1787, forks 216 (observed 2026-08-28T04:05:36.245143+00:00)

## What it is
DreamTalk is the official implementation of a diffusion-based framework for generating expressive, audio-driven talking head videos from a single portrait image. It combines a denoising network, a style-aware lip expert, and a style predictor to produce high-quality talking faces across diverse speaking styles, languages, and out-of-domain portraits.

## Use cases
- generate a talking head video from a portrait and audio clip
- animate a face photo to match speech audio
- create lip-synced avatar videos with different speaking styles
- make a portrait sing along to a song
- research expressive talking head generation with diffusion models
- transfer head pose and speaking style from a reference video to a new face

## When to choose
- you need audio-driven expressive face animation from a single portrait
- you want research-grade talking head generation with style control via 3DMM parameters
- your inputs include songs, multilingual speech, or noisy audio
- you are doing academic research on diffusion-based face generation

## When to avoid
- you need production deployment - checkpoints require an email request restricted to academic research
- you want a plug-and-play tool without GPU setup and 3DMM preprocessing pipelines
- you need real-time or low-latency avatar generation
- you require commercial usage rights

## Facets
- artifact type: library
- maturity: maintenance
- function: deep-learning, video-processing, audio-processing, machine-learning
- domain: computer-vision, deep-learning, artificial-intelligence, media
- platform: python
- tags: talking-head-generation, diffusion-models, face-animation, audio-driven, lip-sync, research-code, video-generation, linux, gpu

## Member repositories
- ali-vilab/dreamtalk (main) score 26

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:36.245143+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 2026-08-30T03:23:28.442085+00:00, confidence not recorded.
  - readme: https://github.com/ali-vilab/dreamtalk (fetched 2026-08-28T04:05:36.245143+00:00, sha 7fbc690daff1)
  - homepage: https://dreamtalk-project.github.io/ (fetched 2026-08-29T11:02:22.309445+00:00, sha 8be9d4a8ec5a)
- Data as of 2026-08-30T08:39:29.467469+00:00.