Recording Stopped After Returning to Yihu? Check First
Back in the app? Check saved audio, transcript state and the destination draft before recording again. Try Yihu’s no-mic audio preview, then review privacy and App Store setup.
Preview privacy before you try
The linked preview uses preset material, needs no login or microphone, and keeps edits only in the open page. Reloading resets edits. Optional read-aloud uses the browser’s voice service; usage analytics require consent.
Keep private originals in approved tools. In the app, on-device recognition and the destination’s storage, sync and access rules are separate steps. Review the installed version before using a source-described workflow. Read the data-flow guide
US listing verified; availability depends on your storefront. No Mac or self-signing for store installation. Bring your workspace address and account; the optional speech model is about 950 MB.
Check saved material and the destination before recording again
On returning to Yihu, inspect saved clips, pending text and errors, then check the current input and existing workspace draft. Keep the original and review prior handoff before deciding to record or send again. A permission dialog’s inactive state differs from true background; resumed note transcription and restarted recording must be checked separately.
Different from audio recovery and format preflight
Recovery concerns route/engine notifications; permission/sample review concerns authorization and format. Here after lifecycle changes, which request remains valid and which work can continue is the task. No setting change, recording, background inference daemon or forced-rebind switch is supplied.
Foreground-return checks
| Event/state | Current implementation boundary | Review on return | Does not establish |
|---|---|---|---|
| Preparing permission dialog | Exception when not background, preparing and requesting permission | Valid request, actual consent and phase | Every inactive cancels, or every return is a new recording |
| True background | Invalidates cancellable short-speech requests, including late permission responses | Old request is not a newly authorized session | Late granted revives the canceled capture |
| Other inactive transition | Original short-speech cancellation, subject to cancellable state | Canceled, complete and unresolved content separately | Cancellation recalls prior insertion/server effects |
| applicationDidBecomeActive | Readiness check and possible idle model warm-up | Actual UI phase and valid pending preparation | Callback establishes capture/recognition success |
| Note transcription pause/resume | Shell pauses outside active, resumes when active | Pending text, errors and actual note state | Background model inference or uninterrupted audio everywhere |
| Audio input/pipeline | System route may have changed | Current permission/format/recovery conditions | Every accessory session seamlessly rebound |
| Page/draft target | Existing insertion/target guards remain | Workbench, page, draft and authority | Same-looking view restores Agent memory or resend eligibility |
| Existing handoff | One-handoff and uncertain-stop boundaries remain | Target content, attempted state and receipts | Foreground clears prior effects or permits blind retry |
Cancellation invalidates short-speech work through session identity/task cleanup; it does not erase effects already committed. Note transcription and live capture are different state machines. One callback cannot establish lock-screen continuity for all paths.
Synthetic return review
Before: Short speech awaiting permission, no final result.
Transition: True background invalidates the old request.
Return: Even with system granted, inspect current state; do not attach late consent to old work.
Next: Explicit choice through the existing entry point; check target/prior content, no automatic rerecord/resend.
This is not a real permission dialog or device regression result. A dialog causing inactive only can preserve pending preparation. Identify the event rather than assuming a returned screen reveals the path.
Rebinding cannot bypass conditions
Current route, format, model and destination retain existing checks. Do not force old converters, ignore failures or adopt a cloud service to make old recording appear continuous. Renaming/replacing drafts does not erase reservations.
When state/content is uncertain, stop within original scope and inspect evidence. No lock-screen/call/accessory/OS matrix was tested, and no zero-loss, latency or background-continuity guarantee is supplied.
FAQ
Does inactive always mean true background?
No. Permission dialogs can cause inactive; preparing permission requests have an exception. Identify actual event and phase.
Can late granted after background revive an old request?
Current short-speech logic invalidates cancellable work on true background. Late permission is not evidence the canceled capture resumes.
Does resumeTranscription mean microphone capture continues?
No. Transcription, audio capture and handoff differ. A method name does not guarantee continuous recording across every path.
Does active prove model recognition completed?
No. Readiness/warm-up is not returned final text. Inspect actual state and errors rather than inferring results from callbacks.
Is there a forced seamless-rebind or lock-screen recording switch?
None is added. This is state/binding preflight, not universal device/OS recovery evidence.
Does foreground clear handoff history so retry is safe?
No. Prior attempts/effects still need review. Lifecycle transitions do not cancel one-handoff/uncertainty rules.
Reading and source
Audio recovery · Permission/sample contract · Target/context review · Handoff claim review · Start · Privacy
Evidence: short-speech lifecycle/cancellation, shell transcription lifecycle, original target guards.
