babelscribe logo

babelscribe

Free, offline speech-to-text and subtitles on any GPU

babelscribe turns any audio or video file into subtitles (SRT, VTT) and text, in any of Whisper's 99 languages, on your own graphics card — AMD Radeon, NVIDIA, Intel Arc or Apple Silicon — or the CPU. No CUDA, no cloud, free and open source (MIT). It runs OpenAI Whisper through whisper.cpp with Vulkan, CUDA or Metal.

GitHubPyPIClaude Desktop extension

Install and run

pip install babelscribe
babelscribe interview.mp4              # auto language, best GPU -> interview.srt
babelscribe interview.mp4 --accurate   # slower, fewest errors
Any GPUAMD Radeon and Intel Arc via Vulkan, NVIDIA via Vulkan or CUDA, Apple via Metal. Picks your discrete card automatically.
99 languagesAuto-detects the language. Thai and Hindi get community fine-tunes in accurate mode.
Works in AI appsMCP server for Claude Desktop, Claude Code, Codex, Antigravity, Gemini CLI, Cursor and VS Code.
PrivateEverything runs on your computer. Nothing is uploaded.

Use it from an AI app

Claude Desktop: download babelscribe.mcpb and double-click it. Then ask: "make subtitles for the newest video in my Downloads". Other apps:

uvx babelscribe mcp

Accuracy (Google FLEURS, 50 utterances per language)

LanguageDefault (turbo)--accurate
English6.1%5.7%
Spanish4.1%4.2%
Japanese (CER)6.4%5.7%
Chinese (CER)6.3%5.3%
Thai (CER)15.9%8.9%
Hindi28.4%12.6%
Vietnamese10.8%9.1%
Arabic11.0%10.6%

Full 16-language table and methodology in the README.

FAQ

How do I run Whisper on an AMD GPU on Windows?

pip install babelscribe, then babelscribe video.mp4. It downloads a Vulkan build of whisper.cpp that runs on AMD Radeon, Intel Arc and NVIDIA cards on Windows and Linux — no ROCm, no CUDA, no compiling.

How do I make SRT subtitles from a video for free and offline?

babelscribe video.mp4 -f srt writes video.srt next to the video. Nothing is uploaded.

What is the most accurate free speech-to-text for Thai?

babelscribe video.mp4 -l th --accurate — Pathumma Whisper (NECTEC) for the text plus Whisper turbo for timing: 8.9% CER on FLEURS vs 15.9% for plain Whisper turbo.

Can Claude or Codex transcribe a video on my computer?

Yes — add babelscribe as an MCP server. The AI app can find a file, transcribe it on your GPU, then proofread, translate or summarise the text.

How fast is it?

About 20x real time with large-v3-turbo on an AMD Radeon RX 9070 XT — a 19-minute talk in 48 seconds.

Guides

ภาษาไทย

ถอดเสียงและทำซับไตเติลจากวิดีโอหรือไฟล์เสียงฟรี ใช้การ์ดจอ AMD / NVIDIA / Intel / Apple ในเครื่องตัวเอง ไม่ต้องอัปโหลด ภาษาไทยแม่นที่สุดด้วย --accurate (Pathumma Whisper ของ NECTEC) หรือใช้ผ่าน Claude Desktop ได้โดยไม่ต้องพิมพ์คำสั่ง

MIT licence · Privacy · built on whisper.cpp and OpenAI Whisper · fine-tunes belong to their authors