Installer whisperkit-cli avec Homebrew

Liens du projet

• Page d'accueil

Commandes et alias

• whisperkit-cli

Historique du projet

Argmax announced WhisperKit in beta after founding the company in November 2023, explaining that production demand for Whisper inference motivated its first open-source project. The beta was released under the MIT license and focused on letting Swift developers run Whisper in apps with minimal setup.

The current Argmax Open-Source SDK Swift repository was created on January 26, 2024. Its README now presents WhisperKit as one part of a broader on-device speech SDK, alongside SpeakerKit for diarization and TTSKit for text-to-speech.

WhisperKit's niche is Apple-platform specialization. Instead of being a universal C/C++ runtime like whisper.cpp, it leans into Swift, Core ML, Apple Silicon, and on-device execution for apps and command-line workflows.

Historique d'adoption

The project grew around the broader shift from cloud-only speech APIs to local transcription for privacy, latency, and cost reasons. The repository reported more than 6,200 stars and more than 570 forks via GitHub's API on July 2, 2026, enough to make it a visible Apple-native Whisper implementation.

Argmax's own materials position WhisperKit as a developer SDK rather than just a demo. The CLI gives package-manager users a direct way to run the same stack from the terminal for transcription tests, batch workflows, and model evaluation.

Modes d'utilisation

Typical CLI use is to install `whisperkit-cli`, inspect transcription options, and run local audio through WhisperKit models. The Hugging Face model page for `argmaxinc/whisperkit-coreml` explicitly points users to installing the CLI with Homebrew and running `whisperkit-cli transcribe --help`.

In the package ecosystem, WhisperKit CLI is most attractive to macOS users who want Apple Silicon acceleration and Swift/Core ML alignment rather than the broad portability of whisper.cpp.

Pourquoi les passionnés de paquets s'y intéressent

WhisperKit CLI represents the Apple-native branch of the local Whisper family: same broad speech-recognition lineage as OpenAI Whisper, but packaged for Core ML and on-device Apple hardware.

Its significance is less about being the oldest Whisper command-line tool and more about making the Apple Neural Engine/Core ML path accessible as a normal installable CLI.

Chronologie

• 2023-11: Argmax says it was founded to help deploy commercial-scale inference workloads on user devices.
• 2024-01-26: The current Argmax open-source Swift repository was created.
• 2024: Argmax open-sourced WhisperKit beta under the MIT license.
• 2025: Argmax model documentation pointed to an ICML 2025 WhisperKit paper and presentation.
• 2026-07-02: GitHub API metadata showed the Argmax Swift repository above 6,200 stars.

Related projects

• OpenAI Whisper supplies the model lineage.
• whisper.cpp is the best-known cross-platform C/C++ Whisper runtime.
• SpeakerKit and TTSKit are sibling packages in the Argmax Open-Source SDK Swift repository.

Sources

• Argmax Open-Source SDK Swift repository: https://github.com/argmaxinc/argmax-oss-swift
• Argmax WhisperKit announcement: https://www.argmaxinc.com/blog/whisperkit
• Argmax repository GitHub API metadata: https://api.github.com/repos/argmaxinc/argmax-oss-swift
• Homebrew formula page: https://formulae.brew.sh/formula/whisperkit-cli
• WhisperKit Core ML model page: https://huggingface.co/argmaxinc/whisperkit-coreml