# Fantasy-AMAP/fantasy-talking

[ACM MM 2025] FantasyTalking: Realistic Talking Portrait Generation via Coherent Motion Synthesis

Repository: https://github.com/Fantasy-AMAP/fantasy-talking
Canonical: https://ross.abutalabs.com/products/fantasy-talking
Homepage: https://fantasy-amap.github.io/fantasy-talking/
Language: Python
License: Apache-2.0
License Family: permissive
Topics: diffusion, diffusion-models, diffusion-transformer, talking-head
Last push: 2026-01-26T17:59:26+00:00

## Health v2 (maintenance only)
Score: 48/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 64, release rhythm 35, longevity 37
- inputs: {"age_days": 518, "days_push": 219, "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 1628, forks 128 (observed 2026-08-28T04:05:13.312196+00:00)

## What it is
FantasyTalking is a research codebase and model for generating realistic talking portrait videos from a single image and an audio clip, built on the Wan2.1 video diffusion transformer with audio-visual alignment. It provides inference scripts and pretrained weights for lip-synced, motion-controllable avatar generation.

## Use cases
- generate a talking head video from a photo and audio
- make a portrait lip-sync to speech
- animate a static face with realistic expressions and body motion
- control expression and motion intensity of an avatar
- research audio-driven portrait video generation
- integrate talking-head generation into ComfyUI workflows

## When to choose
- you need high-fidelity, research-grade talking portrait generation from a single image
- you want controllable motion intensity and identity preservation beyond simple lip-sync
- you have GPU resources and want to build on Wan2.1 diffusion transformer models

## When to avoid
- you need a lightweight real-time lip-sync tool for production apps
- you lack a GPU or cannot download multi-gigabyte base models
- you need training code or a polished end-user application rather than inference scripts

## Facets
- artifact type: library
- maturity: active
- function: video-processing, machine-learning, deep-learning, audio-processing, llm-inference
- domain: deep-learning, computer-vision, artificial-intelligence
- platform: python
- tags: talking-head, diffusion-models, diffusion-transformer, lip-sync, audio-driven-avatar, portrait-animation, acm-mm-2025, research-code, video, audio, linux, gpu

## Member repositories
- Fantasy-AMAP/fantasy-talking (main) score 48

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:13.312196+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 2026-08-30T03:48:19.111969+00:00, confidence not recorded.
  - readme: https://github.com/Fantasy-AMAP/fantasy-talking (fetched 2026-08-28T04:05:13.312196+00:00, sha 7613f83913d6)
  - homepage: https://fantasy-amap.github.io/fantasy-talking/ (fetched 2026-08-29T11:20:56.635838+00:00, sha 02706622488b)
- Data as of 2026-08-30T08:39:29.467469+00:00.