WhisperKit (Argmax OSS SDK)

Use it when

• An iOS or macOS app must transcribe user audio without uploading recordings to any server, for privacy or offline operation
• You need local speech-to-text on Apple silicon with speaker labels, by combining WhisperKit transcription with SpeakerKit diarization or the --diarization CLI flag

What it solves

Not the fit when

• A cross-platform open SDK: the documented open package targets Apple platforms, while Android support is commercial Argmax Pro.
• Treating every local server transport as equivalent: the open CLI documents an HTTP server, while WebSocket streaming is described with Pro features.
• Speaker labels without the companion SpeakerKit integration or CLI diarization option.
• Small installs or instant first run: model assets must be downloaded before local inference.
• cloud-scale batch transcription of huge archives
• android support in the open-source sdk
• windows or linux native apps

Install

brew install whisperkit-cli for the command line, or add the Swift package https://github.com/argmaxinc/argmax-oss-swift in Xcode / Package.swift and select the WhisperKit product (SpeakerKit and TTSKit ship in the same package); requires macOS 14+ and Xcode 16+

Invoke

let pipe = try await WhisperKit(); let results = try await pipe.transcribe(audioPath: "audio.wav") -- models auto-download from HuggingFace; CLI: swift run argmax-cli transcribe --model-path <model> --audio-path <file> (add --stream for microphone, --diarization for speakers), or run the bundled OpenAI-compatible local server with argmax-cli serve

Alternatives

No reviewed alternatives recorded yet.