# Alibaba-Quark/LiveAvatar

[ECCV 2026 Oral] Implementation of "Live Avatar: Streaming Real-time Audio-Driven Avatar Generation with Infinite Length"

Repository: https://github.com/Alibaba-Quark/LiveAvatar
Canonical: https://ross.abutalabs.com/products/liveavatar
Language: Python
License: Apache-2.0
License Family: permissive
Last push: 2026-08-24T15:10:10+00:00

## Health v2 (maintenance only)
Score: 61/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 99, release rhythm 35, longevity 19
- inputs: {"age_days": 272, "days_push": 9, "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 2386, forks 281 (observed 2026-08-28T04:06:42.637772+00:00)

## What it is
LiveAvatar is an open-source implementation of an ECCV 2026 paper for streaming, real-time, infinite-length audio-driven avatar video generation. It uses a 14B-parameter diffusion model with block-wise autoregressive processing to achieve 45 FPS on multi-GPU H800 setups.

## Use cases
- generate a talking avatar video from audio in real time
- stream infinite-length avatar video driven by speech
- build a real-time digital human presenter
- run audio-driven talking head generation on GPUs
- create live streaming avatars for virtual hosts
- research streaming diffusion video generation

## When to choose
- you need real-time, long-running audio-driven avatar video generation
- you have multi-GPU H800-class hardware and want 45 FPS streaming inference
- you want to reproduce or build on a state-of-the-art talking avatar research model

## When to avoid
- you only have a single consumer GPU or CPU-only environment
- you need simple offline talking-head video without streaming requirements
- you want a polished end-user application rather than a research codebase

## Facets
- artifact type: library
- maturity: active
- function: video-processing, machine-learning, deep-learning, llm-inference, speech-recognition
- domain: artificial-intelligence, deep-learning, computer-vision, gpu-computing
- platform: python
- tags: talking-head-generation, audio-driven-avatar, diffusion-model, streaming-video-generation, real-time-inference, eccv-2026, digital-human, video, audio, linux, gpu

## Member repositories
- Alibaba-Quark/LiveAvatar (main) score 61

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:42.637772+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 2026-08-30T02:34:35.643202+00:00, confidence not recorded.
  - readme: https://github.com/Alibaba-Quark/LiveAvatar (fetched 2026-08-28T04:06:42.637772+00:00, sha 6e4a2532ddf1)
- Data as of 2026-08-30T08:39:29.467469+00:00.