Whisper

Model

OpenAI's open-source speech recognition model that runs on your machine

Price: Free, open source
Access: None, runs locally

Repo
Model card

About

Whisper is OpenAI's MIT-licensed speech recognition model and Python package. It transcribes speech in many languages, translates it into English and detects the language, fully offline. It needs ffmpeg, and the large model wants about 10 GB of GPU memory.

What you can do with it

Transcribe audio files offline on your own machine or server
Translate speech in other languages into English text
Generate SRT or VTT subtitles from video or podcast audio

Get started

Install ffmpeg, e.g. brew install ffmpeg
Install the package with pip install -U openai-whisper

Example

import whisper

model = whisper.load_model("turbo")
result = model.transcribe("audio.mp3")
print(result["text"])

Details

Hosting: Self-hosted, Runs locally
Available in: Worldwide
Official SDKs: Python
MCP server: None