Whisper
openai/whisperReviewed by Radar
What it does
Whisper is OpenAI's speech recognition model. It turns spoken audio into text in many languages, can detect which language is spoken, and can translate speech into English. You run it by typing a command in a terminal or from Python code, and it runs on your own computer once a model is downloaded. You need Python and ffmpeg; a graphics card makes it much faster, especially for the larger models, which need about 5–10 GB of video memory. Accuracy varies a lot from one language to another.
How easy it is to install
C · install 1 or 2 tools · Windows
There is no installer: you install Python and ffmpeg (the README suggests Chocolatey or Scoop on Windows), run pip install -U openai-whisper, and then use it by typing commands.
Numbers on GitHub
- Stars
- 111,654on Oct 31, 2026
- Downloads
- —all releases
- Growth
- +310stars in 5 days
Stars over time
Facts
- Runs on
- Windows, macOS, Linux
- You need to install
- Python, FFmpeg
- To build from source
- Python, Graphics card, FFmpeg
- Disk space
- 900 MB–4.2 GB · estimate
- Accounts or keys
- None
- Cost
- Free · Free. The README says both the code and the model weights are under the MIT license.
- License
- MIT (Permissive)
- Last update
- Aug 31, 2026
Risk signals (2)
- Infono release in over a year
The latest GitHub release is v20250625 (2025-06-26); the last commit was on 2026-08-31.
AI review
- Infomay need rust to install
README.md: "You may need `rust` installed as well, in case tiktoken does not provide a pre-built wheel for your platform."
AI review
Evidence is shown in its original language.
What Radar could not check
- How large the model download is and where it is stored on first use
- How slow it is on a computer without a graphics card
- How accurate it is for your language and your recordings (the README shows large differences between languages)
- Whether pip installs a graphics-card-enabled version of PyTorch on Windows without extra steps