Context
Model loading on Android is a silent wait (2–8 s cold start depending on model). iOS v1.7.1 (#144) added a full-screen ModelLoadingOverlay: shimmer animation + cycling status text, mic taps cleanly gated while the model loads, overlay reused during onboarding, and it stays up until real work is observed (no premature dismissal).
What to do
- Compose overlay shown while whisper.cpp / sherpa-onnx context is initializing (app side and IME side)
- Gate the mic button against the loading state (visual disabled state, no dropped taps into the void)
- Reuse in onboarding model download/prewarm step
- Dismiss only when the engine is actually ready
Acceptance criteria
iOS reference: dictus-ios v1.7.1 release notes, PR #164.
Context
Model loading on Android is a silent wait (2–8 s cold start depending on model). iOS v1.7.1 (#144) added a full-screen
ModelLoadingOverlay: shimmer animation + cycling status text, mic taps cleanly gated while the model loads, overlay reused during onboarding, and it stays up until real work is observed (no premature dismissal).What to do
Acceptance criteria
iOS reference: dictus-ios v1.7.1 release notes, PR #164.