中文说明

Turn any video into ready-to-share subtitles. Transub extracts audio with ffmpeg, runs Whisper to transcribe the speech track, and hands the text to an LLM so you get well-translated subtitles without leaving the terminal.

• Overview
• Key Features
• Installation
  • 1. Prerequisites
  • 2. Install Transub
  • 3. Install a Whisper Backend
  • 4. Configure Transub
  • 5. Run the Pipeline
• Configuration Overview
• CLI Cheatsheet
• Development
• Project Layout
• License

Transub orchestrates a reproducible pipeline:

1. Extract audio from a video with ffmpeg.
2. Transcribe speech via Whisper (local, mlx, whisper.cpp, or API).
3. Translate subtitle batches with JSON-constrained prompts.
4. Emit .srt or .vtt files with tuned line breaks and timing.

Intermediate state is cached so interrupted runs can resume without repeating earlier steps.

• End-to-end pipeline — transub run <video.mp4> handles extraction → transcription → translation → export.
• Multiple transcription backends — choose local Whisper, mlx-whisper, whisper.cpp, or OpenAI-compatible APIs.
• Reliable translations — JSON-constrained prompts, retry logic, and configurable batch sizes.
• Subtitle polishing — punctuation-aware line splitting, timing offsets, and optional spacing tweaks when different scripts appear in the same line.
• Stateful execution — cached progress in the work directory (defaults to ~/.cache/transub) avoids rework across runs.

• Python 3.10+
• ffmpeg: Must be installed and available in your system's PATH.
  • Windows: winget install Gyan.FFmpeg or choco install ffmpeg
  • macOS: brew install ffmpeg
  • Linux: sudo apt update && sudo apt install ffmpeg (Debian/Ubuntu) or sudo pacman -S ffmpeg (Arch)

Using uv (Recommended)

uv is a fast Python package installer and resolver. It installs CLI tools in isolated environments.

uv tool install transub

To update later, run:

uv tool upgrade transub

transub supports multiple Whisper backends. Choose one based on your needs:

• Cloud API (Recommended for quick start):

  • Uses OpenAI's Whisper API or compatible endpoints
  • No local installation required
  • Set OPENAI_API_KEY environment variable
  • Configure with backend = "api" during setup

• Local backends (for offline use or custom models):

  • For most users (local, CPU/GPU):

    pip install openai-whisper

  • For Apple Silicon (macOS):

    pip install mlx-whisper

  • For whisper.cpp: Follow the whisper.cpp installation instructions to build the main executable and make it available on your PATH.

Run the interactive setup wizard to create your configuration file.

transub init

The wizard will guide you through selecting the backend, model, and LLM provider for translation.

Note on API Keys: If you use OpenAI for both transcription (Whisper API) and translation (GPT models), they share the same OPENAI_API_KEY by default. If you need separate keys for different services, you can customize api_key_env in the config file for each service.

transub run /path/to/your/video.mp4

Subtitles are written alongside the source video unless you set pipeline.output_dir in your config. Override the cache location with --work-dir when you need an alternate workspace.

Runtime settings live in transub.conf (TOML). Key sections:

• [whisper] — backend selection, model name, device overrides, and extra arguments.
• [llm] — translation provider/model, temperature, batch size, retry policy.
• [pipeline] — output format, line-length targets, timing trim/offset, punctuation and spacing options.

Example:

[pipeline]
output_format = "srt"
translation_max_chars_per_line = 26
translation_min_chars_per_line = 16
normalize_cjk_spacing = true
timing_offset_seconds = 0.05

Run transub configure for an interactive editor, or update the file manually. Configuration files are user-specific and should not be committed.

transub run demo.mp4 --config ~/transub.conf --work-dir /tmp/transub  # override work dir (defaults to ~/.cache/transub)
transub show_config
transub init --config ./transub.conf   # rerun the setup wizard
transub configure                      # edit config (0 saves, Q discards)
transub run demo.mp4 --transcribe-only # export raw transcription only
transub run demo.mp4 -T              # short flag for transcribe-only
transub --version                    # print the installed version

The work directory (defaults to ~/.cache/transub) stores audio, transcription segments, translation progress, and pipeline state. If a run is interrupted, re-running the same command resumes where it left off. Use --work-dir to point at a custom cache location when needed.

If you want to contribute to transub, you can set up a development environment.

1. Clone the repository:

   git clone https://github.com/PiktCai/transub.git
   cd transub

2. Create and activate a virtual environment:

   python3 -m venv .venv
   source .venv/bin/activate

3. Install in editable mode with development dependencies:

   pip install -e ".[dev]"

4. Install a Whisper backend for testing:

   pip install openai-whisper

python -m unittest

• Source lives in transub/ (cli.py, config.py, transcribe.py, translate.py, subtitles.py, etc.).
• Add tests beside related modules (e.g., transub/test_subtitles.py).
• Use Rich console utilities and transub.logger.setup_logging for consistent output.

transub/
├── audio.py
├── cli.py
├── config.py
├── subtitles.py
├── transcribe.py
├── translate.py
└── test_subtitles.py

This project is distributed for personal use and study; there is no formal contribution process at this time.
Transub is released under the MIT License.