# Soul-AILab/SoulX-FlashHead

SoulX-FlashHead: A unified 1.3B-parameter framework designed for high-fidelity, infinite-length, and real-time streaming portrait video generation.

Repository: https://github.com/Soul-AILab/SoulX-FlashHead
Canonical: https://ross.abutalabs.com/products/soulx-flashhead
Homepage: https://soul-ailab.github.io/soulx-flashhead
Language: Python
License: Apache-2.0
License Family: permissive
Last push: 2026-05-28T03:47:22+00:00

## Health v2 (maintenance only)
Score: 53/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 84, release rhythm 35, longevity 14
- inputs: {"age_days": 209, "days_push": 97, "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 1011, forks 117 (observed 2026-08-28T04:03:13.224713+00:00)

## What it is
SoulX-FlashHead is a 1.3B-parameter framework for high-fidelity, infinite-length, real-time streaming talking-head portrait video generation driven by audio. It includes inference code, pretrained checkpoints, a Gradio demo, and the 782-hour VividHead training dataset.

## Use cases
- generate talking head videos from audio
- real-time streaming digital human avatar
- audio-driven lip sync video generation
- build a real-time virtual presenter
- infinite-length portrait video generation
- low-latency digital human interaction

## When to choose
- you need real-time streaming talking-head generation on consumer GPUs
- you want high-fidelity audio-driven portrait video with no identity drift over long sequences
- you need a lightweight model (96 FPS on a single RTX 4090) for digital human applications

## When to avoid
- you need full-body or multi-person video generation
- you lack a CUDA GPU, as inference requires NVIDIA hardware
- you need training code, which is not yet fully released

## Facets
- artifact type: library
- maturity: active
- function: video-processing, machine-learning, deep-learning, llm-inference, speech-recognition
- domain: deep-learning, artificial-intelligence, computer-vision
- platform: python
- tags: talking-head-generation, digital-human, audio-driven-video, streaming-inference, lip-sync, avatar-generation, diffusion-distillation, comfyui, video, audio, gpu, linux, docker

## Member repositories
- Soul-AILab/SoulX-FlashHead (main) score 53

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:13.224713+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 2026-08-30T07:11:49.040497+00:00, confidence not recorded.
  - readme: https://github.com/Soul-AILab/SoulX-FlashHead (fetched 2026-08-28T04:03:13.224713+00:00, sha 26a293abaab8)
  - homepage: https://soul-ailab.github.io/soulx-flashhead (fetched 2026-08-29T13:11:37.787456+00:00, sha 835a2b88f4b7)
- Data as of 2026-08-30T08:39:29.467469+00:00.