Whisper
Whisper is OpenAI's general-purpose speech recognition model, released with both code and weights under the MIT licence. It transcribes many languages, translates speech to English and detects the spoken language in one network, in six sizes from 39 million to 1.55 billion parameters. It installs with pip and runs as a command line or a Python library, and it is the model most other open-source transcription tools build on.
What does Whisper do?
Whisper is "a general-purpose speech recognition model" that does "multilingual speech recognition, speech translation, and language identification" in one network. The README lists six sizes: tiny (39M), base (74M), small (244M), medium (769M), large (1550M) and turbo (809M), with English-only variants of the four smaller ones. Turbo, the default, is described as an optimised large-v3 with "faster transcription speed with a minimal degradation in accuracy".
The README gives its own approximate memory needs: about 1 GB of VRAM for tiny and base, 2 GB for small, 5 GB for medium, 10 GB for large and 6 GB for turbo. Audio is processed in a sliding 30-second window. Word timestamps, speaker labelling and streaming are not described in the README; the projects built on Whisper, listed below, add them.
Key facts
- Licence: MIT on both the code and the model weights, stated in the README's licence section.
- Six sizes: tiny (39M), base (74M), small (244M), medium (769M), large (1550M) and turbo (809M), with English-only variants of the four smaller ones.
- The README's approximate VRAM needs: about 1 GB for tiny and base, 2 GB for small, 5 GB for medium, 10 GB for large, and 6 GB for turbo.
- Turbo, the default, is described as an optimised large-v3 with faster transcription and minimal loss of accuracy; it is not trained for translation.
- Audio is processed in a sliding 30-second window.
- Install: pip install -U openai-whisper, plus ffmpeg, with install steps for Ubuntu, Arch, macOS and Windows; Python 3.8 to 3.11 stated as compatible.
- Speed figures in the README were measured on an A100 GPU; CPU and Apple Silicon are not discussed, which is where whisper.cpp comes in.
- Latest tagged release when read: v20250625 on 2025-06-25.
What does it replace, and where does it fall short?
Whisper replaces the Deepgram and AssemblyAI transcription APIs for recorded audio where the audio has to stay on your own systems. The licence is the reason it anchors this category: "Whisper's code and model weights are released under the MIT License."
Where it falls short: it is a model and a command-line tool, with no HTTP server, no streaming and no speaker labels of its own. For those, see whisper.cpp (server, CPU and Apple Silicon), WhisperX (word timestamps and speakers) and Speaches (OpenAI-compatible server) on this site.
How does Whisper run?
pip install -U openai-whisper, then the whisper command or the Python API. It needs ffmpeg, and the README gives install steps for Ubuntu, Arch, macOS and Windows. Python 3.8 through 3.11 are stated as compatible. The README's speed figures were measured on an A100 GPU; CPU and Apple Silicon are not discussed, which is where whisper.cpp comes in. OpenAI does not offer a hosted version of this repository; its transcription API is a separate product.
Who is Whisper for?
A team that wants the reference model with the plainest licence, on a GPU, for batch transcription. A team that needs live audio should look at Moonshine, Vosk or sherpa-onnx.
What limits does the README state?
"The turbo model is not trained for translation tasks" and returns the original language even when asked to translate. "Whisper's performance varies widely depending on the language." The English-only models "tend to perform better" for English at the tiny and base sizes.
Questions people ask
Is Whisper open source?
Yes. The README states that Whisper's code and model weights are released under the MIT licence, which is OSI approved and permits commercial use. The training data itself is not published, but the label on this site only requires OSI licences on code and weights.
Can I use Whisper commercially?
Yes. MIT permits commercial use, modification and redistribution of both the code and the weights with attribution. The many products built on it, from whisper.cpp to the desktop dictation apps on this site, do exactly that.
Does Whisper run on a CPU or a Mac?
The reference package is built around PyTorch and its README quotes speeds on an A100 GPU; it does not discuss CPU or Apple Silicon. For those, whisper.cpp is the port to use: quantised models, Metal and Core ML on Apple Silicon, and CPU-only inference.
How does Whisper compare with Deepgram or AssemblyAI?
Whisper gives you the model to run yourself with no per-minute bill and no audio leaving your network, but no hosted API, no streaming and no speaker labels of its own. The alternatives to Deepgram page places Whisper, whisper.cpp, WhisperX and the streaming engines by need.
Which Whisper size should I use?
The README makes turbo the default for English and multilingual transcription, notes that the English-only .en models tend to perform better at the tiny and base sizes, and warns that turbo returns the original language even when asked to translate. Pick by the VRAM you have, from about 1 GB for tiny to about 10 GB for large.
Sources
- Whisper README and MIT LICENSE: github.com/openai/whisper, read 2026-09-16.
Compared with the others
On the open-source alternatives to Deepgram page, Whisper is the pick for the reference model on a gpu. MIT code and weights, six sizes, multilingual recognition plus translation to English. The model every other entry here builds on or competes with.
Also on that page: whisper.cpp for a cpu, a mac, or a phone, WhisperX for word timestamps and speaker labels, Speaches for an openai-compatible server.
On the open-source alternatives to AssemblyAI page, Whisper is the pick for the reference model on a gpu. MIT code and weights, six sizes, multilingual recognition plus translation to English. The model every other entry here builds on or competes with.
Also on that page: WhisperX for word timestamps and speaker labels, Speaches for an openai-compatible server, whisper.cpp for a cpu, a mac, or a phone.
More speech-to-text engines
A dependency-free C and C++ port of Whisper under MIT that runs on CPUs, Apple Silicon and phones, with quantised models, an HTTP server and bindings for a dozen languages.
Batched Whisper transcription under BSD-2-Clause with word-level timestamps from forced alignment and speaker labels from pyannote, the closest open-source match to a hosted transcription API's output.
NVIDIA's speech models and toolkit: Apache-2.0 code, and the Parakeet TDT 0.6B v3 model under CC-BY-4.0 with 25 European languages, automatic language detection, punctuation and word timestamps.
A runtime under Apache-2.0 from the next-generation Kaldi team that runs speech to text, text to speech, speaker diarization and voice activity detection locally on CPUs, NPUs and phones, with no internet connection.
An offline speech recognition toolkit under Apache-2.0 with streaming, models of about 50 MB, 20+ languages, a reconfigurable vocabulary and speaker identification, from small boards to clusters.
An on-device speech toolkit under MIT built for live streaming, with speech-to-text models trained from scratch in sizes down to 1 MB, one library across Python, the browser, phones and desktops.
Added September 16, 2026. Every claim above comes from the project's README, LICENSE or model card, read on September 16, 2026, or from the GitHub API on the date shown in the panel. Found an error? Write to reveneau@licheo.com and it is fixed in the next weekly pass. Repository: github.com/openai/whisper.
Want Whisper running on your own servers?
Reveneau sets it up on your hardware, connects it to your product, and builds the software around it.