Repolode
Submit
zackees

zackees/transcribe-anything

CLI tool that transcribes local files or URLs with Whisper, supporting multiple backends and speaker-labeled output.

Open-source alternative to:Otter.ai

View on GitHubWebsite
Stars
1,483
Forks
146
Open issues
15
Last commit
12 days ago

What is it?

transcribe-anything is a command-line front end for Whisper speech-to-text. Give it a local audio or video file, or a YouTube or other supported URL, and it downloads and normalizes the audio, then writes plain text, SRT and VTT subtitles, and JSON output. Each backend runs in its own isolated environment that is built on first use, so CUDA and torch dependencies stay out of your main Python install.

A `--device` option selects the backend. Options include the original openai-whisper CPU and CUDA path, insanely-fast-whisper for batched GPU transcription with an optional FlashAttention2 variant, WhisperX for VAD, word alignment and diarization, SenseVoice for multilingual and CJK-focused recognition, Intel Arc XPU support, and an MLX backend for Apple Silicon. Diarization-capable backends produce a speaker.json file that groups text by speaker.

The project also offers a Python API, optional subtitle burn-in into video, and a daemon mode that runs a shared HTTP transcription server with optional token auth.

Key features

  • 01Transcribes local files and YouTube or other URLs into TXT, SRT, VTT and JSON outputs.
  • 02Selects among multiple Whisper backends with --device, auto-detected if omitted.
  • 03Builds each backend in an isolated environment on first use.
  • 04Produces speaker.json with speaker-labeled text from diarization-capable backends.
  • 05Offers a Python API and an optional daemon mode for shared transcription servers.

Use cases

  • Generating subtitles for YouTube videos or local recordings.
  • Transcribing interviews or meetings with per-speaker labels.
  • Running private, offline-style transcription on your own GPU or Apple Silicon Mac.
  • Batch transcription on a shared GPU host through the daemon mode.

Quick start

  1. 1

    Install the package from PyPI

    pip install transcribe-anything
  2. 2

    Transcribe a YouTube video in CPU mode

    transcribe-anything "https://www.youtube.com/watch?v=dQw4w9WgXcQ"

Things to know

  • The first run of each backend downloads its own dependencies, about 10 GB for --device insane.
  • Some backends are platform-specific: --device mlx needs Apple Silicon, and --device xpu is experimental.
  • Speaker diarization with --device whisperx needs a HuggingFace token.

In collections

CLI & terminal toolsAI video & audio tools

Topics

#speech-to-text#whisper#transcription#subtitles#diarization#cli#python

FAQ

What is transcribe-anything?
CLI tool that transcribes local files or URLs with Whisper, supporting multiple backends and speaker-labeled output. transcribe-anything is a command-line front end for Whisper speech-to-text. Give it a local audio or video file, or a YouTube or other supported URL, and it downloads and normalizes the audio, then writes plain text, SRT and VTT subtitles, and JSON output. Each backend runs in its own isolated environment that is built on first use, so CUDA and torch dependencies stay out of your main Python install.
Is transcribe-anything free to use? What license does it have?
transcribe-anything is open source under the MIT license. Check the license terms before using it in commercial products.
Is transcribe-anything still maintained?
Actively maintained. The last commit to transcribe-anything was on Oct 26, 2026.
What can I use transcribe-anything for?
Generating subtitles for YouTube videos or local recordings. Transcribing interviews or meetings with per-speaker labels. Running private, offline-style transcription on your own GPU or Apple Silicon Mac. Batch transcription on a shared GPU host through the daemon mode.
How do I install transcribe-anything?
Install the package from PyPI: pip install transcribe-anything Transcribe a YouTube video in CPU mode: transcribe-anything "https://www.youtube.com/watch?v=dQw4w9WgXcQ"
Where can I get transcribe-anything?
The source code and installation instructions are on GitHub: https://github.com/zackees/transcribe-anything

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects