Video Process

Japanese video/audio processing pipeline with transcription, translation, and subtitle generation.

Features

• 🎙️ Transcribe Japanese audio using OpenAI GPT-4o
• 🌏 Translate Japanese to Chinese using X.AI Grok-3
• 📝 Generate aligned bilingual subtitles with stable-ts (Whisper large-v3)
• 🎬 Create videos with embedded subtitles using ffmpeg

Quick Start

# Process all media files in input directory
bash main/main.sh ./main/input ./main/output

# Debug mode with verbose output
DEBUG=1 bash main/main.sh ./main/input ./main/output

Requirements

Python Packages

pip install openai stable-whisper numpy numba

External Tools

• ffmpeg
• stable-ts

API Keys

• OpenAI API key (configure in main/scripts/transcribe.py)
• X.AI API key (configure in main/scripts/translate_x.py)

Supported Formats

• Video: mp4, avi, mkv, mov, webm
• Audio: mp3, wav

Output Structure

For each processed file, outputs are created in <output_dir>/<filename>/:

• audio.wav - Extracted audio
• ja_lines.txt - Transcribed Japanese text
• translations.json - Chinese translations
• aligned_ja.srt - Time-aligned Japanese subtitles
• subtitles.srt - Final bilingual subtitles
• final_with_subtitles.mp4 - Video with embedded subtitles

Manual Subtitle Editing

After processing, you can manually edit subtitles:

1. Edit subtitles_editable.srt in the output directory
2. Copy back: cp subtitles_editable.srt subtitles.srt
3. Re-encode video: bash ffmpeg_command.txt

License

MIT