WhisperSubTranslate

WhisperSubTranslate is a free desktop app for creating translated SRT subtitles from video speech. It's for people who want to subtitle videos on their own computer, with offline transcription and translation after downloading the models.

• Speech recognition: Generate subtitles with whisper.cpp using large-v3-turbo by default, or choose smaller models for less powerful hardware.
• Subtitle timing repair: Use large-v2 Sync or Sync Lite when ordinary speech models produce subtitles that drift out of sync.
• Offline translation: Translate subtitle text with Tencent's Hy-MT2 1.8B or the larger 7B model without an account or API key after the initial download.
• Online translation: Send subtitle text to MyMemory, DeepL, OpenAI, Gemini or Claude, or use a custom OpenAI-compatible endpoint; API keys are required where applicable.
• Job tracking: Queue videos, follow processing progress and keep job history locally.

The app has a portable Windows distribution. Speech recognition can use NVIDIA GPUs through CUDA, Vulkan-capable AMD and Intel GPUs, or the CPU. Local translation also supports GPU acceleration and CPU processing. Hy-MT2 1.8B needs about 2 GB of VRAM or 4 GB of RAM; the 7B model needs about 8 GB of VRAM or 12 GB of RAM.

Video stays on your machine. Hy-MT2 keeps translation text local too, while online translation sends it to the selected provider. The app generates subtitles from audio; it doesn't extract embedded subtitle tracks or read text shown on screen.