Your Voice.
Organized Locally.
An intelligent iOS voice assistant that transcribes speech and extracts tasks, events, and journal notes. Runs completely offline on your iPhone using local models.
Designed for Liquid Intelligence
A stunning UI packed with spatial glassmorphism, responsive offline workflows, and native performance.
WhisperKit Transcription
Transcribe spoken notes instantly with high-fidelity, device-optimized speech models. By executing natively on Apple's Neural Engine, Notiary achieves blazing speeds with zero internet bandwidth required.
100% Private
No external servers, cloud databases, or API keys. Your audio prompts, schedules, and action items never leave your device.
Local LLM Processing
A quantized Qwen model runs locally via llama.cpp. It structures unorganized transcript thoughts into reminders, events, and notes offline.
Intelligent Date Resolution
No external servers or complex timezone configs. Notiary's built-in DateResolver parses absolute, relative ("tomorrow at 2pm"), and clock variations into local system event times without drifting.
See Notiary in Action
Click a preset voice note below to see how local ASR, LLM extraction, and Apple EventKit sync work together.
Select a Voice Note
Click any sample audio command to run the on-device AI transcription & event mapping simulation:
Technical Specifications
Under the hood of our highly optimized Swift and C++ local stack.
| ASR Model | WhisperKit (Apple Neural Engine optimized, fully on-device) |
| LLM Inference | llama.cpp / llama.xcframework C-Interop (Quantized Qwen GGUF) |
| Date Parser | Custom Swift DateResolver (ISO8601, NSDataDetector, keyword relative cascade) |
| Data Integrations | EventKit (Direct Apple Reminders & Apple Calendar Sync) |
| Hardware Support | iOS 17.0+ (Universal Swift Concurrency, CoreDevice, XcodeGen targets) |
