Apple Native Speech Recognition (Local Apple ASR)

Apple Native Speech Recognition utilizes macOS system speech frameworks accelerated by the Apple Neural Engine (ANE) on Apple Silicon (M1/M2/M3/M4) chips. In DuRT, it powers ScreenCaptureKit system internal audio live captions and microphone dictation.

---

Core Characteristics & Boundaries

• 100% On-Device Offline Processing: Audio is processed entirely inside your Mac's secure memory and immediately discarded. Zero network data is ever uploaded;
• Sub-300ms Ultra-Low Latency: Streaming audio produces immediate word segmentation and on-screen caption rendering;
• Zero Cost: No third-party API keys required; permanently free without usage caps;
• Broad Language Support: Dozens of languages maintained directly by Apple (English, Chinese, Japanese, Korean, German, French, Spanish, Italian, etc.).

WARNING

Boundary Notice: Media File Transcription Not Supported Apple's native speech framework is strictly built for live streaming audio. In DuRT, Apple Native Speech is only available in the Realtime Speech panel and cannot transcribe imported MP3/MP4 media files. For file transcription, please use OneASR Gateway or OpenAI ASR.

---

System Permissions & Setup

1. System Permissions:
   • Microphone: Needed for capturing external microphones or AirPods audio;
   • Screen Recording (ScreenCaptureKit): Required to capture internal audio from Zoom, browsers, or media players without kernel driver extensions (like BlackHole).
   • See macOS Permissions Guide.
2. Download Offline Speech Packs (Recommended):
   • Open System Settings ➔ General ➔ Keyboard;
   • Under Dictation, enable the toggle;
   • Select Languages and check your required languages to download offline models.

---

Configuration in DuRT

1. Open DuRT, go to Settings ➔ AI Capabilities ➔ click Realtime Speech;
2. Locate Apple (Built-in) in the list;
3. Ensure it is toggled Enabled;
4. Select your Primary Recognition Language (e.g. en-US or zh-CN).