# microsoft/VibeVoice

Open-Source Frontier Voice AI

Repository: https://github.com/microsoft/VibeVoice
Canonical: https://ross.abutalabs.com/products/vibevoice
Homepage: https://microsoft.github.io/VibeVoice/
Language: Python
License: MIT
License Family: permissive
Last push: 2026-07-24T09:24:29+00:00

## Health v2 (maintenance only)
Score: 60/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 94, release rhythm 35, longevity 26
- inputs: {"age_days": 373, "days_push": 40, "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 53225, forks 6004 (observed 2026-08-28T04:12:18.017609+00:00)

## What it is
VibeVoice is Microsoft's open-source frontier voice AI framework combining a next-token diffusion text-to-speech model for expressive, long-form, multi-speaker conversational audio (up to 90 minutes, 4 speakers) with VibeVoice-ASR, a unified speech-to-text model for hour-long transcriptions with speaker and timestamp annotations. It is distributed as Python models on Hugging Face and PyPI, with finetuning code and an edge CPU inference engine.

## Use cases
- generate a multi-speaker podcast from a text script
- synthesize long-form conversational speech with consistent voices
- transcribe hour-long audio with speaker labels and timestamps
- convert text to expressive speech with emotion and singing
- cross-lingual speech generation between Mandarin and English
- run speech recognition on CPU at the edge
- finetune a TTS or ASR model on custom voice data

## When to choose
- you need long-form multi-speaker speech synthesis beyond typical 1-2 speaker TTS limits
- you want open-source, MIT-licensed state-of-the-art TTS or ASR models
- you need structured transcriptions with speaker and timestamp metadata
- you want to finetune speech models on your own data

## When to avoid
- you need a lightweight production TTS API with minimal compute
- you require strict guarantees against voice cloning misuse
- you need real-time low-latency streaming synthesis on modest hardware
- you only need simple single-phrase text-to-speech

## Facets
- artifact type: library
- maturity: active
- function: tts, speech-recognition, machine-learning, llm-inference
- domain: speech-processing, artificial-intelligence, deep-learning
- platform: python, cross-platform
- tags: text-to-speech, speech-synthesis, asr, multi-speaker, podcast-generation, diffusion, voice-ai, long-form-audio, audio, gpu

## Member repositories
- microsoft/VibeVoice (main) score 60

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:12:18.017609+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 2026-08-29T16:19:34.204180+00:00, confidence not recorded.
  - readme: https://github.com/microsoft/VibeVoice (fetched 2026-08-28T04:12:18.017609+00:00, sha ba850794c6d6)
  - homepage: https://microsoft.github.io/VibeVoice/ (fetched 2026-08-28T18:07:25.099994+00:00, sha a784683f4008)
  - registry_pypi: https://pypi.org/pypi/vibevoice/json (fetched 2026-08-28T18:07:25.109002+00:00, sha b7b8a783e96d)
- Data as of 2026-08-30T08:39:29.467469+00:00.