json2srt

Convert OpenAI Whisper JSON output with word-level timestamps to precise SRT subtitle files.

Purpose

OpenAI Whisper generates JSON files with word-level timestamps when using --word_timestamps True. This tool converts those JSON files to SRT subtitle format, preserving per-character/word timing accuracy rather than sentence-level timestamps.

This is particularly useful for scenarios requiring precise synchronization:

• Language learning applications
• Precise subtitle production
• Audio analysis and research

Usage

1. Generate JSON file with word-level timestamps using Whisper:

   whisper audio.mp3 --word_timestamps True --output_format json

2. Place generated JSON file in same directory as program
3. Run program to auto-convert to precise SRT files

Features

• ✨ Precise timestamps: Maintains Whisper's word-level timing precision
• 🚀 Batch processing: Auto-processes all JSON files in directory
• ⚡ Concurrent processing: Uses multi-core CPU for faster processing
• 📝 BOM support: Generates SRT files with BOM for correct Chinese character display
• 🔍 Smart filtering: Auto-skips Whisper noise markers (like [_TT_123])

Run

go run main.go

Or compile and run:

go build -o json2srt main.go
./json2srt

Dependencies

• Go 1.16+

Output

• Each JSON file generates corresponding SRT file with BOM
• SRT file timestamps are precise to each word's start and end time

License

MIT