Argmax Open-Source SDK
A Swift package of on-device inference frameworks for Apple platforms covering speech-to-text, speaker diarization, and text-to-speech.
The Argmax Open-Source SDK bundles three separate library products in one Swift package: WhisperKit for speech-to-text built on OpenAI's Whisper, SpeakerKit for speaker diarization built on Pyannote, and TTSKit for text-to-speech built on Qwen-TTS, or an umbrella ArgmaxOSS product that imports all three at once. WhisperKit includes memory-efficient loading for large audio files and a local server with a documented API; TTSKit supports custom voices, real-time streaming playback, and style instructions on its larger model; SpeakerKit supports RTTM output and combining diarization with transcription. Each kit also ships its own Swift command-line tool.
A separate commercial Argmax Pro SDK extends the same kits with real-time transcription with speakers, custom vocabulary, a local server for non-native apps, and Android support.
It requires macOS 14 and Xcode 16 or newer, and is distributed under the MIT license.