Whisperkit Coreml is an open-source model hosted on Hugging Face for automatic speech recognition. It forms the foundation of WhisperKit, an on-device speech AI SDK developed for Apple Silicon hardware.
The model supports transcription of audio files through a command-line interface available via Homebrew on macOS. Users install the whisperkit-cli tool and then run commands such as transcribe with an audio path to process local files. Options allow selection of specific model variants including the base model or combinations such as large-v3 with a distil prefix, along with flags for verbose output and other inference settings. The repository also links to example notebooks on Google Colab and Kaggle for further experimentation.
WhisperKit itself is presented as part of the broader Argmax OSS project, which focuses on on-device speech AI for Apple Silicon. A related paper and presentation from ICML 2025 provide additional technical context on the implementation. The model card indicates it is quantized and intended for use with the WhisperKit library or the provided CLI.
In the Voice, TTS & speech space, Whisperkit Coreml takes a focused approach. It focuses on enabling accurate, on-device speech-to-text transcription using open-source models optimized for Apple Silicon devices. It is built as an open-source project for developers building speech recognition apps for Apple devices. Whisperkit Coreml is open source under the MIT license. It runs on the command line, and it can be self-hosted.
It is developed by Argmax, and the product first shipped in 2024. Development happens publicly on GitHub with 6.3k stars and 12 commits in the last 90 days. Key capabilities include speech recognition, on-device inference, and Core ML support.
Latest indexed changes and source events
argmaxinc/whisperkit-coreml verified by the PulseGate indexer
Other apps tracked under the same category.
