# meizhong986/WhisperJAV

ASR/STT subtitle generator. Uses Qwen3-ASR, local LLM, Whisper, TEN-VAD. Noise-robust for JAV

Repository: https://github.com/meizhong986/WhisperJAV
Canonical: https://ross.abutalabs.com/products/whisperjav
Homepage: https://github.com/meizhong986/WhisperJAV/releases/latest
Language: Python
License: MIT
License Family: permissive
Topics: aitranslate, hallucination, japanese, llm, modelscope, qwen3, qwen3-asr, speech-to-text, speechenhancement, subtitling, ten-vad, whisper, zipformer
Last push: 2026-08-22T12:55:42+00:00

## Health v2 (maintenance only)
Score: 94/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 98, longevity 75
- inputs: {"age_days": 1050, "days_push": 11, "days_rel": 18, "gap_med": 6, "n_releases_24m": 38}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2170, forks 184 (observed 2026-08-28T04:06:21.529990+00:00)

## What it is
WhisperJAV is a local, privacy-preserving subtitle generator for Japanese adult videos, combining Qwen3-ASR, Whisper, TEN-VAD/FireRedVAD voice activity detection, and a local LLM for translation. It addresses known ASR failure modes on noisy, long-form audio through scene-based segmentation, VAD clamping, and defensive decoding.

## Use cases
- generate japanese subtitles for long-form videos
- transcribe noisy audio with lots of non-speech sounds
- prevent whisper hallucination on silence and breathing
- translate japanese speech to subtitles with a local llm
- batch subtitle generation without uploading media to the cloud
- segment audio at acoustic scene boundaries before ASR

## When to choose
- you need noise-robust Japanese ASR for adult or otherwise acoustically difficult content
- you want fully local processing with no cloud upload of media
- you need subtitle files with accurate timing from long recordings

## When to avoid
- you need general-purpose ASR for clean studio-quality speech
- you want a cloud-hosted or managed transcription service
- you need languages other than Japanese

## Facets
- artifact type: application
- maturity: active
- function: speech-recognition, nlp, machine-learning, llm-inference, audio-processing, cli
- domain: speech-processing, media, artificial-intelligence
- platform: python, cli, cross-platform, self-hosted
- tags: asr, stt, subtitling, whisper, qwen3-asr, vad, japanese, noise-robust, subtitle-generation, hallucination-mitigation, natural-language-processing, localization

## Member repositories
- meizhong986/WhisperJAV (main) score 94

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:21.529990+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 2026-08-30T02:49:19.755395+00:00, confidence not recorded.
  - readme: https://github.com/meizhong986/WhisperJAV (fetched 2026-08-28T04:06:21.529990+00:00, sha 01470b0ccea2)
  - homepage: https://github.com/meizhong986/WhisperJAV/releases/latest (fetched 2026-08-29T10:29:25.153367+00:00, sha 376ccd641fc0)
- Data as of 2026-08-30T08:39:29.467469+00:00.