Argmax Open-Source SDK

The Argmax Open-Source SDK is a collection of on-device speech AI frameworks for Apple Silicon, written in Swift. It bundles WhisperKit for speech-to-text, SpeakerKit for speaker diarization and TTSKit for text-to-speech. The repository has about 6,400 stars on GitHub, is MIT-licensed, and is actively maintained with updates in 2026.

Overview

Argmax’s SDK runs speech models entirely on iPhone and Mac hardware using Core ML, so transcription and synthesis work offline with no cloud round-trip. WhisperKit ports OpenAI’s Whisper models to Apple Silicon with model variants tuned for speed or accuracy; SpeakerKit uses Pyannote models to label who spoke when; TTSKit generates speech with Qwen-TTS models, including custom voices and streaming playback. The package installs through Swift Package Manager, ships a command-line tool, and can even serve an OpenAI-compatible local audio API for existing clients.

Key Features

• WhisperKit: on-device speech recognition with OpenAI Whisper models
• SpeakerKit: speaker diarization powered by Pyannote models
• TTSKit: text-to-speech with Qwen-TTS, custom voices and streaming
• Optimized for Apple Silicon with Core ML, fully offline inference
• Swift Package Manager installation plus a ready-made CLI
• Local server mode implementing the OpenAI Audio API
• Model gallery with downloadable variants for different devices

💳 Pricing

The open-source SDK is free under the MIT license. Argmax also sells a Pro SDK with additional models and advanced features such as real-time transcription with speaker labels, custom vocabulary and Android support, so the overall offering follows a freemium model.

Who Is It For

The Argmax Open-Source SDK is for iOS and macOS developers who want private, offline speech recognition, diarization or synthesis inside their apps without paying per-minute cloud transcription fees.