# kleinlee/DH_live

每个人都能用的数字人

Repository: https://github.com/kleinlee/DH_live
Canonical: https://ross.abutalabs.com/products/dh_live
Language: Python
License Family: other
Last push: 2026-08-16T11:07:33+00:00

## Health v2 (maintenance only)
Score: 67/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 98, release rhythm 35, longevity 54
- inputs: {"age_days": 764, "days_push": 17, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2131, forks 436 (observed 2026-08-28T04:06:17.219392+00:00)

## What it is
DH_live (mini) is an open-source 2D talking-head digital human toolkit that generates real-time lip-synced avatar video from a single reference video, with no per-person training required. It includes video data preparation, model checkpoints for face detection, mesh landmarks, video generation, and speech features, plus web and realtime conversation demos (VAD-ASR-LLM-TTS-avatar) that can run on CPU and even directly in mobile browsers.

## Use cases
- create a talking avatar from a single video without training
- run a real-time digital human conversation pipeline with ASR, LLM and TTS
- deploy a lightweight digital human that runs in a mobile browser
- build a web server serving a digital human avatar
- generate lip-synced video offline from audio
- embed a digital human assistant in an app or mini-program

## When to choose
- you need an ultra-lightweight talking-head avatar with very low compute per frame
- you want a no-training, out-of-the-box digital human from one reference video
- you need real-time avatar inference on CPU or mobile devices
- you want a full realtime voice conversation pipeline with an avatar

## When to avoid
- you need high-resolution, studio-quality avatar rendering
- you require 3D avatars or full-body animation
- you need a commercially licensed product (the repo has no license file)
- you need offline video synthesis on Linux or macOS (Windows only)

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, video-processing, speech-recognition, tts, chatbot, web-framework, llm-inference
- domain: artificial-intelligence, computer-vision, speech-processing, web-development, chatbots
- platform: python, cross-platform, windows
- tags: digital-human, talking-head, lip-sync, real-time-avatar, video-avatar, no-training, lightweight-inference, browser-inference, cpu-inference, gradio, web-server, linux, macos, android, ios

## Member repositories
- kleinlee/DH_live (main) score 67

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:17.219392+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 2026-08-30T02:52:11.058623+00:00, confidence not recorded.
  - readme: https://github.com/kleinlee/DH_live (fetched 2026-08-28T04:06:17.219392+00:00, sha 8cacfc542eba)
- Data as of 2026-08-30T08:39:29.467469+00:00.