Japanese Audio Transcription App

A Python desktop application for Japanese audio transcription using Whisper, with an interactive UI for playback and sentence navigation. Perfect for language learners who want to study audio with timestamps.

Features

• Load any Japanese audio file (.mp3, .wav).
• Transcribe audio into sentences with timestamps using Whisper.
• Play and stop audio.
• Click on sentences to jump to the corresponding point in the audio.
• Status updates for model loading and transcription progress.
• Safe on macOS with PyTorch + multiprocessing.
• Highly sentence-based transcription for better navigation.
• Translate transcripts to Vietnamese (optional).

Screenshots

Videos

Installation

1. Clone the repository:

git clone https://github.com/chuongmep/japanese-audio-transcriber.git
cd japanese-audio-transcriber

2. Create a virtual environment and activate it:

python -m venv venv
source venv/bin/activate   # macOS/Linux
venv\Scripts\activate      # Windows

3. Install dependencies:

pip install -r requirements.txt

requirements.txt should include at least:

PySide6
whisper
torch
pydub
simpleaudio

4. (macOS only) Make sure ffmpeg is installed for audio processing:

brew install ffmpeg

Windows

winget install --id Gyan.FFmpeg -e

Usage

Run the application:

python main.py

1. Click Load Audio to select a Japanese audio file.
2. Click Transcribe to generate sentences with timestamps.
3. Use Play / Stop to listen to the audio.
4. Click on any sentence on the right panel to jump to that point in the audio.

Notes

• Whisper automatically uses the best model available. The current version uses small. You can change to medium or large for higher accuracy.
• The application uses a separate thread for transcription to prevent crashes on macOS with PyTorch.
• For large audio files, transcription may take several minutes.

Contributing

Contributions are welcome! You can improve:

• Word-level clickable transcription.
• Support for other languages.
• Export transcripts to CSV or SRT format.
• UI enhancements.

License

MIT License – see LICENSE file.

Issues

• Whisper TypeError: argument of type 'NoneType' is not iterable

pip install git+https://github.com/openai/whisper.git