LLMVoice

Real-time voice transcription with AI summaries. 100% on-device for complete privacy (Tech Demo - not for production use).

LLMVoice - iOS 26 Speech & AI Technology Demo

A proof-of-concept application showcasing Apple's latest iOS 26 technologies for speech recognition and on-device AI, including MLX Swift for running open-source language models locally.

Technology Demonstration - Not for Production Use

LLMVoice is a technical showcase for developers and early adopters interested in Apple's latest frameworks and on-device AI capabilities.

What This Demo Does

• Real-time voice transcription using iOS 26 speech APIs • AI-powered summarization with multiple model options • Modern SwiftUI interface with streaming updates • 100% on-device processing for complete privacy

Key Technologies

MLX Swift Framework Apple's MLX (Machine Learning Extensions) enables efficient inference of quantized language models on Apple Silicon with optimized GPU acceleration.

Apple Intelligence On supported devices (iPhone 15 Pro or later), use Apple's built-in system models for summarization without downloading additional files.

Open Weights Models via MLX Choose from multiple open-source models running locally: • Qwen2.5 (0.5B) - 150MB, 29+ languages, 32k context • Gemma 3 (1B) - 300MB, 140+ languages, 32k context • Llama 3.2 (1B) - 500MB, 8 languages (EN/ES/FR/DE/IT/PT/HI/TH)

All models use 4-bit quantization for mobile efficiency.

SpeechAnalyzer & SpeechTranscriber (iOS 26) Low-latency, real-time on-device speech recognition using Apple's newest APIs.

Modern Swift & SwiftUI Built with Swift 6, Observable macro, async/await, and actor-based concurrency.

How It Works

1. Select AI model (Apple Intelligence or download MLX model)
2. Tap microphone to start recording
3. Speak naturally - real-time transcription appears
4. Tap stop when finished
5. AI generates concise summary instantly
6. Summaries saved locally, persist between sessions
7. Swipe to delete, tap to view original transcription

Privacy-First

Everything runs on-device: • Audio never leaves your device • Local speech recognition and AI inference • No cloud processing or internet required (except model downloads) • No tracking or data collection

Requirements

• iOS 26.0+ (iPadOS 26.0+ for iPad) • Apple Silicon recommended • Storage for models (150MB - 500MB each) • Microphone and speech recognition permissions

Technical Limitations

• MLX requires real devices (not iOS Simulator) • Models downloaded on first use • Limited customization options • Primarily English-focused • No export/sharing features • Proof of concept - not production-ready

Perfect For

• Developers exploring iOS 26 and MLX Swift • AI enthusiasts interested in on-device LLMs • Students learning SwiftUI and modern concurrency • Anyone curious about running open-weights models on mobile

Development Highlights

Demonstrates MLX Swift integration with Apple frameworks, including model downloading, GPU memory management, streaming inference with real-time metrics, and efficient on-device transformer execution.

Mac Catalyst Compatible

Runs on macOS 26+ with Apple Silicon for unified cross-platform experience.

Note on Model Downloads

First-time use requires downloading your chosen model from HuggingFace. Apple Intelligence is available immediately on compatible devices without downloads.