feat: Biking-aware audio enhancement via Core Motion

[devin-ai-integration]

Summary

When Core Motion detects the user is cycling or in automotive motion, the voice recording pipeline automatically improves capture quality to cut through wind noise:

New file: SpeechPipeline/MotionActivityManager — singleton that wraps CMMotionActivityManager, filters for medium/high confidence cycling/automotive activity, and exposes a reactive isBiking flag + Notification.Name.motionBikingStateDidChange. Bootstrapped from AppDelegate.didFinishLaunching. Graceful no-op if Core Motion is unavailable or permission denied — isBiking stays false and the audio pipeline behaves identically to before.

Audio session voice isolation — both recording paths in MessageBox (startRecording AVAudioRecorder path and beginStreamingRecording AVAudioEngine/Deepgram path) now check MotionActivityManager.shared.isBiking when configuring AVAudioSession:

// before:  mode: .default
// after:   mode: isBiking ? .voiceChat : .default

.voiceChat enables Apple's built-in echo cancellation + noise suppression + AGC.

High-pass wind filter — in the streaming STT (AVAudioEngine) path, when biking is detected an AVAudioUnitEQ node configured as a 200 Hz high-pass filter is inserted between inputNode and mainMixerNode. The tap reads from the EQ's output so captured audio is already filtered before reaching Deepgram. The EQ is detached on teardown. No filter is applied when stationary.

Info.plist — added NSMotionUsageDescription for the Core Motion permission prompt.

Link to Devin session: https://app.devin.ai/sessions/d7cfd1e15cbe426387e3f22a9b053796

[devin-ai-integration]

Original prompt from API User

Implement motion-aware audio improvements for the Loop iOS app — specifically when the user is biking.

Repository: https://github.com/getathelas/LoopHarness (Swift, SwiftUI, iOS/macOS/visionOS app)

#``# Feature: Biking-Aware Audio Enhancement

#``#``# Goal
When Core Motion detects the user is biking (CMMotionActivityManager → automotive/cycling activity), automatically improve audio capture quality for voice recording / speech-to-text.

#``#``# Implementation

1. Core Motion detection:

   • Use CMMotionActivityManager to monitor activity transitions
   • Detect .automotive or .cycling activity states
   • Expose a simple reactive flag (e.g. isInMotion / isBiking) that other parts of the app can observe

2. Voice Isolation auto-enable:

   • When biking is detected, set AVAudioSession category to .playAndRecord with mode .voiceChat
   • Enable Voice Isolation mic mode if available (iOS 16+): AVAudioSession.setMode(.voiceChat) which enables Apple's built-in echo cancellation + noise suppression + auto gain control
   • When motion stops, revert to default audio session settings

3. High-pass filter via AVAudioEngine:

   • If the app uses AVAudioEngine for audio capture, insert an AVAudioUnitEQ node configured as a high-pass filter at ~150-200Hz to cut wind rumble before the audio is sent to speech recognition
   • Only activate this filter when biking/motion is detected
   • Deactivate when stationary

4. Graceful fallback:

   • If Core Motion is unavailable or permission denied, default to standard audio session (no crash, no degradation)
   • Log when motion-aware audio is active vs inactive for debugging

#``#``# Notes

• Look for existing audio session configuration in the codebase and extend it rather than replacing it
• The app uses voice recording via Action Button → speech-to-text pipeline
• Keep changes additive and well-documented
• Add ashbhat as a reviewer on the PR

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment, CI, and merge conflict monitoring

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
loop-harness	Ready	Preview, Comment	Jun 23, 2026 1:21am