# soniqo/speech-swift

AI speech toolkit for Apple Silicon — ASR, TTS, speech-to-speech, VAD, and diarization powered by MLX and CoreML

Repository: https://github.com/soniqo/speech-swift
Canonical: https://ross.abutalabs.com/products/speech-swift
Homepage: https://soniqo.audio
Language: Swift
License: Apache-2.0
License Family: permissive
Topics: apple-silicon, asr, coreml, ios, macos, mlx, neural-engine, on-device, speaker-diarization, speech-enhancement, speech-recognition, speech-to-speech, swift, text-to-speech, tts, voice-activity-detection
Last push: 2026-09-02T17:56:41+00:00

## Health v2 (maintenance only)
Score: 83/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 100, release rhythm 100, longevity 15
- inputs: {"age_days": 211, "days_push": 0, "days_rel": 0, "gap_med": 4.5, "n_releases_24m": 27}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1161, forks 158 (observed 2026-09-03T02:15:18.025884+00:00)

## What it is
An open-source Swift toolkit for on-device speech AI on Apple Silicon, providing ASR, TTS, speech-to-speech, VAD, and speaker diarization via MLX (Metal GPU) and CoreML (Neural Engine). It ships as a Swift package with modular components, plus a Homebrew CLI and an OpenAI-compatible HTTP/WebSocket server, all running fully offline.

## Use cases
- transcribe audio to text locally on a Mac
- generate speech from text on-device
- build a privacy-preserving voice agent with VAD, STT, LLM, and TTS
- run speaker diarization on meeting recordings
- clone a voice and synthesize speech offline
- add dictation to an iOS app without cloud APIs
- serve local speech AI through an OpenAI-compatible realtime endpoint

## When to choose
- you need fully offline, private speech AI on Apple Silicon Macs or iOS
- you want a native Swift async/await API instead of C++ bridges
- you need low-latency realtime transcription or full-duplex speech-to-speech
- you want to avoid cloud API costs and per-minute pricing

## When to avoid
- you need to run on non-Apple hardware (Android/Windows/Linux support is via separate SDKs, not this Swift package)
- you target macOS below 15 or iOS below 18
- you need MLX-based models on iOS, which are macOS-only
- your project is not in the Swift/Apple ecosystem

## Facets
- artifact type: library
- maturity: active
- function: speech-recognition, tts, audio-processing, machine-learning, llm-inference, cli, http-server, websocket
- domain: speech-processing, machine-learning, artificial-intelligence, apple-ecosystem, mobile-development, developer-tools
- platform: cli
- tags: on-device, apple-silicon, mlx, coreml, voice-activity-detection, speaker-diarization, voice-cloning, offline-inference, voice-agents, speech-to-speech, openai-compatible-api, macos, ios, swift, apple-ecosystem

## Member repositories
- soniqo/speech-swift (main) score 83

## Provenance
- Observed fields: from GitHub, fetched 2026-09-03T02:15:18.025884+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T06:33:46.966051+00:00, confidence not recorded.
  - readme: https://github.com/soniqo/speech-swift (fetched 2026-09-03T02:15:18.025884+00:00, sha 9afbe53963bb)
  - homepage: https://soniqo.audio (fetched 2026-08-29T12:38:49.158648+00:00, sha 5ae9b4faa19d)
  - site_page: https://soniqo.audio/getting-started (fetched 2026-08-29T12:38:49.161246+00:00, sha 15a90c94f512)
  - site_page: https://soniqo.audio/faq (fetched 2026-08-29T12:38:49.163119+00:00, sha 877129c3effe)
- Data as of 2026-08-30T08:39:29.467469+00:00.