WhisperSubTranslate offers a free, local desktop application for generating and translating video subtitles. It enables users and developers to create multilingual SRT subtitles from video audio via speech-to-text, with an optional offline translation capability, eliminating cloud reliance, signups, and costs. The core benefit is complete local control over subtitle processing.

How It Works

The app uses whisper.cpp for local speech-to-text transcription to generate SRT files. Translation is handled either offline via bundled Tencent Hy-MT2 models (1.8B default, 7B optional), requiring no API keys or network access, or through online engines (DeepL, OpenAI, Gemini) using user-provided API keys. Automatic model downloads and a no-Python, no-manual-setup user experience simplify deployment.

Quick Start & Requirements

Users download the latest portable archive from Releases, extract, and run WhisperSubTranslate.exe. Developers need Node.js (>= 20.19 or >= 22.12) and can run npm install and npm start. whisper.cpp (with CUDA on Windows) and FFmpeg are included/downloaded. Linux developers require cmake, build-essential, git, ffmpeg. CUDA acceleration requires the NVIDIA CUDA Toolkit.

Highlighted Details

• Features 100% local speech-to-text and optional offline translation for data privacy.
• Bundles Tencent Hy-MT2 models for offline translation, with larger options available.
• Includes specialized "Sync" and "Sync Lite" models to fix subtitle synchronization drift.
• Supports automatic model downloads for speech recognition and translation.
• UI supports Korean, English, Japanese, Chinese, Polish; targets 14 translation languages.
• All data, including settings and API keys, is stored locally.

Maintenance & Community

Contributions are welcomed via pull requests, with guidelines in CONTRIBUTING.md. Community translation is managed on Weblate. Direct contributions are encouraged for project support.

Licensing & Compatibility

Licensed under GPL-3.0. External API usage requires adherence to their terms. The GPL-3.0 license imposes copyleft restrictions, potentially impacting integration with closed-source commercial software.

Limitations & Caveats

The tool does not perform OCR for on-screen text or extract embedded subtitle tracks. Specialized sync models are for synchronization issues only, not general transcription.