About
The Issue
When users play back a recording and tap on a transcript paragraph timestamp, the audio does not play from the correct position. The timestamps are off by a few seconds, so the words being spoken don't match what's displayed in the transcript at that moment.
Expected behavior: Tapping a timestamp should seek to the exact point in the audio where that paragraph begins.
Technical Details
Platform: iOS (SwiftUI)
- Speech Recognition: FluidAudio SDK (on-device ASR with VAD)
- Audio: AVFoundation (AVAudioEngine for recording, AVPlayer for playback)
The root cause is likely latency in voice activity detection (VAD) — the timestamp is captured when speech is detected, not when it actually began in the audio.
Deliverables
- Diagnose the exact cause of the timestamp offset
- Implement a fix so timestamps align correctly with audio playback
- Test across multiple recordings of varying lengths
- Ensure fix works for both new recordings and doesn't break existing functionality
Requirements
- Strong experience with iOS/SwiftUI development
- Familiarity with AVFoundation audio APIs
- Experience with speech recognition or audio processing is a plus
- Ability to debug timing/synchronization issues
Contract duration of less than 1 month.
Mandatory skills: iOS, iOS Development, SwiftUI, Swift, Core ML
Languages
- English
Notice for Users
This job comes from a TieTalent partner platform. Click "Apply Now" to submit your application directly on their site.