WhisperJAV is a Python-based ASR/subtitle tool for Japanese Adult Video content, combining Whisper, Qwen3-ASR, and TEN-VAD with scene-based processing and various backends. It offers GUI and CLI usage, multiple pipelines, translation, and offline model support.
Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (1 so far).
What it is
WhisperJAV is a subtitle generator for Japanese Adult Videos (JAV) that uses multiple ASR backends (Whisper, Qwen3-ASR, anime-whisper) and VAD/scene-based processing to produce subtitles. It includes features for AI translation, ensemble two-pass transcription, and adjustable pipelines and sensitivity settings. The project is implemented in Python and distributed under MIT license.
How it works
The tool provides seven processing pipelines with configurable backends, scene detection, speech enhancement, and speech segmenters. It supports per-pipeline backends like Faster-Whisper, OpenAI Whisper, Kotoba, Qwen3-ASR, and anime-whisper, with VAD segmentation (including TEN for timing). It includes ChronosJAV and Anime-Whisper components, two-pass ensemble merging, and multiple output formats (SRT, VTT). It also features translation options via Ollama, DeepSeek, Gemini, Claude, GPT-4, and local LLMs, plus a local translation path that downloads llama-cpp-python on first use.
Getting started
The README provides Quick Start commands for GUI and CLI:
whisperjav-gui
# Basic usage
whisperjav video.mp4
# Specify mode and sensitivity
whisperjav audio.mp3 --mode balanced --sensitivity aggressive
# Process a folder
whisperjav /path/to/media_folder --output-dir ./subtitles
AI Translation
# Generate and translate
whisperjav video.mp4 --translate
# Or translate existing subtitles
whisperjav-translate -i subtitles.srt --provider deepseek
Local LLM Translation (Legacy)
whisperjav-translate -i subtitles.srt --provider local
Supported Input Formats
Any format FFmpeg can read: MP4, MKV, AVI, MOV, WMV, FLV, WAV, MP3, FLAC, M4A, M4B, and many more.
## Recent releases
Latest releases include:
- v1.8.14 (2026-05-10): Quality Hardening + Bug Fixes; stability and FAQ revamp; ASR preset retune across 3 sensitivities for faster-whisper and openai-whisper; improved spee.
- v1.8.13 (2026-05-05): anime + qwen + ollama improvements; WhisperSeg becomes default for Ensemble / Qwen / Colab; China internet friendly updates.
- v1.8.12 (2026-04-30): WhisperSeg VAD + tight defaults retune; Whisper-VAD-EncDec-ASMR onnx backend used.
- v1.8.11 (2026-04-20): sanitizer hardening; bug fixes for Colab and Ollama translation.
- v1.8.10.post3 (2026-04-09): hotfixes and quality hardening atop v1.8.10.
## Traction
Stars: 2080. Forks: 171. Open issues: 135.
## Behind the repo
Not included in the facts block.
## Caveats
License: MIT. Created: 2023-10-18. Last push: 2026-05-10. Language: Python.