Packages for keyword “on-device”
These packages are available as a package collection, usable in Xcode or SwiftPM.
Qwen3Speech
AI speech toolkit for Apple Silicon — ASR, TTS, speech-to-speech, VAD, and diarization powered by MLX and CoreML
CoreMLLLM
Run LLMs on Apple devices with CoreML, optimized for Apple Neural Engine + GPU
DesertAnt
On-device AI SDKs for iOS, macOS, Android, and the web. Small, focused models that run fully offline in Swift, Kotlin, and JavaScript with Core ML, LiteRT, and WebAssembly.
swift-litert-lm
Gemma 4 on iPhone as a Swift package: one dependency line, five lines of code, ~50 tok/s on iPhone 17 Pro. Metal GPU, text + image + audio, in-app download. Also an Apple Foundation Models backend (iOS 27).
ComposablePipelines
Open, runtime-agnostic Swift package for composing AI pipelines: SwiftUI-like DSL, Codable AST, compiler, and an observable walker. Public mirror from elix-toolchain.
GLiNER2Swift
Swift port of Gliner2 framework
LoudKit
Local text-to-speech: 28 voices, 10 languages, voice cloning, SDKs for Python, Swift, Go, Rust and TypeScript
SynheartCore
iOS/macOS SDK for the Human Stat Index (HSI) — biosignals, behavior, on-device state computation
ContextSDK
Real-world context insights SDK for Swift apps, enhancing user experience and engagement with a privacy-first approach
sudachi-swift
Swift bindings for the Sudachi Japanese morphological analyzer (sudachi.rs) — on-device tokenization for iOS and macOS via UniFFI
AiVoiceKit
On-device voice dictation, command routing, and AI-assisted rewriting for macOS — a standalone, host-app-agnostic Swift package.
Logits
Dependency-free Swift helpers for on-device ML output: softmax, top-k, labelled predictions, bounding-box IoU, letterbox mapping, and non-max suppression. Core ML adapters included.
swift-speech-lanes
On-device, multi-language speech-to-text for Apple platforms. Runs one SpeechAnalyzer lane per locale and returns the highest-confidence transcript.
WatermarkPurge
On-device multi-pass text watermark remover for Apple Silicon (Swift + MLX)
14 packages.