Transcribe Anything
Local speech-to-text software turns files and web videos into text and subtitles, with Whisper backends, Apple Silicon acceleration and optional speaker labels.
Transcribe Anything turns recordings and online videos into text and subtitles using Whisper-based speech recognition. It's a command-line tool with a Python API for people processing media locally or adding transcription to their own applications.
- Files and web videos: Transcribe local media or YouTube and Rumble URLs into raw text and subtitle files.
- Backend choice: Choose Whisper, insanely-fast-whisper or WhisperX, with an optional FlashAttention2 backend for supported NVIDIA CUDA systems.
- Timing and speaker labels: WhisperX provides word-level timestamps, alignment and optional speaker identification; speaker.json groups transcript passages by speaker.
- Apple Silicon transcription: The MLX backend supports multiple languages and custom vocabulary prompts for names or specialized terms.
The project is free and open source under the MIT license. Local transcription processes audio on your machine. NVIDIA GPU acceleration is available on Windows and Linux, while Apple Silicon uses MLX. The CUDA FlashAttention2 backend doesn't run on macOS.
Optional pyannote speaker labeling requires your own Hugging Face token and acceptance of the model's terms. A community-maintained RunPod Serverless deployment offers cloud GPU processing instead of local execution.
