whisperkit-coreml

whisperkit-coreml

Holds Core ML speech recognition models for WhisperKit, which turns audio into text on Apple Silicon without a server. Made for Apple app developers who want transcription that stays on the device.

Made by argmaxinc on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face.

About

What it is: whisperkit-coreml is a set of Core ML speech recognition models used by WhisperKit. WhisperKit belongs to Argmax OSS, an open kit for running speech AI directly on Apple Silicon. The models turn recorded speech into text on the device, so no cloud service is needed. The files carry the MIT license.

Who it suits: It suits developers who build Mac or other Apple Silicon apps and want local transcription, and anyone curious about speech recognition that runs on their own hardware.

Industry | Who uses it | How they use it | Example
Software | Apple app developers | Add speech to text to an app | Let a notes app transcribe a recording
Productivity | Tool makers | Transcribe meetings on the device | Turn a call recording into text
Accessibility | Caption builders | Make text from spoken audio | Add captions to a local video
Privacy | Careful teams | Keep audio off outside servers | Transcribe sensitive interviews locally

If you build for Windows or Linux, you can skip it, since the kit targets Apple Silicon.

Where to use it: It runs on Apple Silicon through the Argmax OSS Swift project. You can download the files from any computer with Python, but running them needs the Apple side.

• Apps built with the Argmax OSS Swift SDK
• Apple Silicon computers
• Local transcription with no server in between

Use case | Who | What they get back
Transcription | App developers | Text from a spoken recording
On-device captions | Media makers | Caption text created locally
Research | Speech engineers | A Core ML speech model to test and compare

How to set it up: You need an Apple Silicon computer for running and Python for the download tool. The card does not list the model sizes inside, so check the file list on the model page before you fetch everything.

1. Install the Hugging Face hub tools.
2. Download the model files.
3. Add the Argmax OSS Swift SDK to your app and point it at them.

pip install -U huggingface_hub
hf download argmaxinc/whisperkit-coreml

How to use it: Use the Argmax OSS Swift SDK to load the models and give it audio. The card shows no run command or code, so the exact calls come from that project's own pages.

Piece | Role
WhisperKit | The speech recognition toolkit that uses these models
Argmax OSS | The open kit that runs speech AI on Apple Silicon
Argmax Pro SDK | A separate option for live transcription with speakers and custom words

Example: You add the SDK to a Swift app, point it at these model files and give it a voice recording. The app returns the spoken words as text. The card shows no sample output, so results depend on your audio.

For newcomers and experts: On-device means the audio is processed on your own computer, not sent to a server. Start with the SDK's own examples before building your own app.

The card points to a paper and a talk about WhisperKit from the ICML conference, which cover the method. It also says the Pro SDK is the route for real-time transcription with speaker labels and custom vocabulary.

Get an Apple Silicon computer
Download the model files
Add the Argmax OSS SDK to your app
Feed audio and read the text

Good to know: This card is short, so rely on the Argmax OSS project for setup and usage details.

• Pro: live transcription with speakers and custom vocabulary sits in the separate Argmax Pro SDK.
• Paper: WhisperKit has a paper and a talk from the ICML conference.
• License: MIT.
• Cost to run: none named for these models. The Pro SDK is a separate product.
• Data it reaches: the audio you give it, handled on your own device.
• Not for: Windows or Linux speech apps, since the kit is made for Apple Silicon.

On Hugging Face, 13 November 2026

• Task: automatic speech recognition
• Library: whisperkit
• Downloads in the last 30 days: 60,717,190
• Likes: 266
• License: mit
• Tags: whisperkit, coreml, whisper, quantized, automatic-speech-recognition

Links

• Model: https://huggingface.co/argmaxinc/whisperkit-coreml
• Files: https://huggingface.co/argmaxinc/whisperkit-coreml/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download argmaxinc/whisperkit-coreml

Then follow the usage in its model card. Nothing in this kit runs by itself.