Audio Processing Standards Skill
This skill provides the authoritative instructions for developing and maintaining the audio recording and processing pipelines of TranscribeJS.
1. Dual Recording Path Integrity
The application uses two separate recording paths in App.tsx because MediaRecorder is unreliable inside Tauri's Linux webview.
| Condition | Path | Implementation |
|---|---|---|
tauriEnv && linuxEnv | Native | Rust cpal via invoke('start_native_recording') / invoke('stop_native_recording'), returns WAV bytes |
| Everything else | Web | navigator.mediaDevices.getUserMedia() + MediaRecorder API, produces webm/ogg blob |
Critical Rules
- Rule: Any change to recording logic must be tested on both paths. A fix that works on web may break Tauri native, and vice versa.
- Rule: Do not unify the two paths unless both are fully validated.
2. Environment Configuration
- Rule: The dev server (
dev.ts) must setCross-Origin-Opener-Policy: same-originandCross-Origin-Embedder-Policy: require-corpheaders. These are required forSharedArrayBufferandOfflineAudioContext. - Rule: Wrap all async audio service calls (API, FS, Processing) in
try/catch. - UI Feedback: Always reflect audio processing errors in the UI (set
statusto'error').
3. Services
- Location:
src/services/audio/. - Logic: Audio duration, conversion, and splitting.
