Voice recording UX: pause, resume and save

Put the current state beside the timer

TapNote displays Paused below an elapsed time, with Delete and Save as separate actions. Quiz AI also labels Paused and adds a maximum-duration message near its controls. Neither asks the person to interpret the timer alone as proof that capture is active.

Use a persistent word or clearly announced state in addition to color and motion. A frozen timer might mean pause, a stalled process or an inactive screen. Recording and Paused should be mutually exclusive states with matching controls. When capture has not yet started, avoid displaying the same active treatment that appears after microphone access succeeds.

If the product has a duration limit, show it before the final seconds and explain the boundary. Does reaching it save the captured material, pause the session or stop further input? Do not silently discard the recording at a limit. The exact limit is a product decision; the recorded maximum in Quiz AI is not a recommendation for every recorder.

Label the action that will happen next

Voice Recorder & Transcribe combines a Paused label with Bookmark and Save around a central recording control. Kiwi AI shows Paused and a resume label beside its controls. The visible word matters more than whether the central button is red, round or visually prominent.

A paused recording needs a control named Resume, not another ambiguous Pause label. If an icon changes meaning with state, update its accessible name at the same moment. A play triangle can mean resume capture or play the existing audio, so use surrounding copy to distinguish them. Review the label and implementation together, not as separate design tasks.

Keep finalization distinct from resuming. Save should not unexpectedly continue capture, and Resume should not create a second file unless the product explains a new segment. If the session contains several captured segments separated by pauses, preserve their order and show one coherent duration. The person should not need to understand audio-file assembly to finish a note.

Keep elapsed audio and contextual notes on one timeline

Voice Recorder Audio Memos shows RESUME below a waveform and a precise timer. Smart Noter places Comments beside Audio and shows a timestamped comment while recording controls remain available. A recorder can support detail without letting metadata replace the essential capture state.

Decide whether the main duration represents captured audio or wall-clock session time. Pauses make those values diverge. Bookmarks and comments should attach to a defined audio position, not to whichever number happened to be displayed when the annotation was entered. If wall-clock timestamps are also useful, label them separately.

Allow a note to be added without losing the recording controls or dismissing an unsaved session. Keep the audio segment identifier and position with that note. A photo attachment or keyboard opening should not silently restart the timer. After returning from the attachment picker, show the current capture state again rather than assuming the person remembers it.

Use session history when interruptions need explanation

TLDL shows a Timeline containing Recording Started and Paused events, plus Bookmark, Add Photo and Add Note. VoiceNotes uses a simpler screen with a large timer and explicit Pause and Save controls. One provides a history of the session; the other prioritizes the immediate decision.

A timeline is useful when people need to understand gaps or organize a long session. It should distinguish recorded content from pause events rather than suggesting silence was captured throughout. Keep event labels concise and let the user reach the corresponding audio when that operation is supported. Do not manufacture a transcript entry for time when capture was off.

For a simpler recorder, retain enough state to explain an interruption even if no timeline is visible. Returning from a call, permission change or app restart should not leave a convincing recording animation over an inactive microphone. Show what was preserved and what needs to happen next. A less decorative but truthful state is preferable to an uninterrupted-looking display.

Review stop controls and labels in the paused state

Summary shows PAUSED with Cancel, a central stop-shaped control and Resume. Noteflow shows RECORDING PAUSED while a play-shaped control is labeled Pause. That mismatch is a useful review case: the visible state and the control label can disagree even when the surrounding layout looks familiar.

Name Stop according to its consequence. If it finishes capture and opens a review screen, say Finish recording where space permits. If it merely pauses, use Pause. If saving starts transcription or consumes an allowance, distinguish preserving the audio from requesting the extra processing. Do not make the user choose between losing the source and committing to an unclear service.

Use Cancel carefully. Cancelling a modal can mean keeping the session available, while cancelling capture can mean discarding it. A destructive exit should explain what remains and provide a safe return. Closing an attachment picker should never discard the underlying recording. Route every exit through the same preservation policy rather than relying on the container’s default dismiss behavior.

Preserve captured segments before starting extra processing

Model permission pending, ready, recording, paused, finalizing, saved and failed states explicitly. Keep captured segments and a stable session identity through pause and resume. Persist completed segments frequently enough for the product’s recovery requirements. A successful UI transition cannot substitute for confirmation that the audio data exists.

Separate recording finalization from upload, transcription and summarization. The source can be saved locally while later processing is pending. If processing fails, retry it against the preserved audio instead of asking the user to record again. Prevent accidental repeated Save taps from creating duplicate jobs or duplicate notes.

Handle microphone denial, lost input routes, incoming calls and storage pressure as real states. Do not promise background recording until the platform implementation has been tested. Explain when capture has stopped and make the preserved portion findable. A waveform can support feedback, but test actual recorded sound rather than treating the rendered shape as a validity check.

Listen to the saved result after each interruption

Record a phrase, pause, speak another phrase, resume and speak a third. Save the result and listen to it. The middle phrase should be absent if Paused means capture is off, while the first and third should remain in order. Compare bookmark positions and displayed duration with that actual file.

Repeat with an incoming interruption, a changed audio input, denied permission and limited storage. Test leaving the app and returning under the supported recording policy. Measure lost or corrupt sessions, finalization failures and recovery success. A tap on Save is not a reliable completion event if the file cannot be reopened.

Check large text, screen-reader names and one-handed access to Pause and Resume. Distinguish capture errors from transcription errors in analytics. Avoid uploading private audio simply to debug a timer mismatch when state and duration metadata would suffice. The primary success is a usable preserved recording with a state the person understood throughout.

Review capture and finalization as one flow

Use these checks on a working build, not only on the polished reference state.

• ✓Is permission success distinct from the appearance of a recording screen?
• ✓Do Recording, Paused and the next-action label agree?
• ✓Does the duration represent a defined clock, with annotations using the same audio timeline?
• ✓Are pause, finish, save and discard different understandable actions?
• ✓Can interruptions preserve captured segments without claiming capture continued?
• ✓Can transcription or upload retry against already saved audio?
• ✓Does the final file reopen and contain the expected segments?

Connect live capture to the rest of the note

The AI note-taking app review covers source material, transcripts and outputs. Live capture needs its own state contract before those later features begin.

Compare mobile loading states when separating finalization, upload and transcription. They are different operations, not one indefinite spinner.

The mobile error-state examples help keep the preserved audio and a useful next action available after a processing failure.