Argmax OSS Swift

Argmax OSS Swift is a speech AI SDK for developers building iOS and macOS apps that process audio on the user's device. Its separate libraries cover transcription, identifying speakers and generating spoken audio.

• Speech recognition: WhisperKit transcribes recordings and microphone audio with OpenAI Whisper, including multilingual models and custom fine-tuned models in Core ML format. It can process long recordings without loading the entire file into memory.
• Speaker identification: SpeakerKit uses Pyannote to distinguish speakers and identify when each person speaks in a recording.
• Speech generation: TTSKit runs Qwen3-TTS locally, with built-in voices, multiple languages and streaming playback. Its larger, macOS-only model also accepts instructions for speaking style.
• Local audio API: The included server lets apps use OpenAI SDK clients for transcription and translation into English, with streamed results and word or segment timestamps. It returns JSON rather than plain text or subtitle files.

The SDK is open source under the MIT license and targets Apple Silicon. You can use the libraries separately or together, and a command-line app supports use outside a Swift app. Speech inference runs locally; default model downloads use Hugging Face, and TTSKit caches its models on the device after download.

The package requires macOS 14 or later. TTSKit requires macOS 15 or iOS 18 or later, and its larger model runs only on macOS.