Use this workflow when feedback is surprising, unclear, or especially useful:
C-e H f groups current and recent feedback actions. x explains the captured item, c opens a guided change, v prepares a voice mapping, t tunes the named voice used there, and e prepares an exact cue remap. h opens recent feedback, whose changes use the selected frozen record. Tuning a named voice affects its other uses; a mapping changes which matching items use a voice.
Opening an editor from this menu does not bypass its preview, matching criteria, scope or Apply/Save steps. On return, the menu selects the same entry. A guided draft can be resumed later in this session, while leaving a modified rule list still asks before discarding it. If the captured source item has been replaced, prepare the change again; the editor does not silently apply it to a different target.
When retained feedback matches the source buffer, position, and unchanged text, the report describes the exact output queued, including the source and resolved resources. It does not reinterpret old output with your latest settings. Without an exact record, the report is labelled as a simulation of what the current rules would produce at point.
A simulation uses captured text and context. Play simulation plays that baseline, not a recording of earlier sound. It supports field review and draft previews but is never added to Recent Feedback. A prefix argument with C-e E selects a simulated occasion explicitly. q returns to the source or menu item from which the report was opened.
Expand a field and choose Edit named voice bolden, or its reported voice name, to open the voice editor. Multiple named voices receive separate links. This edits the entry in the reported palette and affects its uses there. Change field instead changes the field’s feedback rule. Leaving the voice editor returns to the same link; opening it does not save or apply anything.
Press C-e H f h for recent feedback. Each row preserves one recorded presentation within the history’s size limit. Its Content column reads the preview with captured voice annotations, including changes within the entry; other columns and titles use the normal interface voice. P requests replay with the engine and physical voice recorded for the original delivery.
Open Feedback Details for the selected presentation.
Replay retained feedback with recorded voices without adding history.
Audition only its auditory cues.
Prepare a voice remap from the selected record.
Replace, suppress, or restore one exact earcon.
Change the bounded history limit for this session.
Include or exclude feedback from Aural interfaces themselves.
To change the same policy from a menu, open C-e H f H. Select Retained records and use RET for an exact nonnegative limit, or Shift-left/right to adjust by one. Lowering the limit immediately discards the oldest excess records; zero clears and disables history. Increasing it cannot restore those records. The menu’s i toggles interface recording, and w saves both defaults through Customize. Retained feedback itself is never saved by this action.
Aural interface feedback is excluded by default so it does not displace the feedback being investigated. This includes their prompts, completion choices, and opening/closing sounds. Ordinary completion elsewhere remains eligible. Changing the setting affects new recordings, not entries already retained.
Feedback Details selects a separate read-only buffer and gives a brief introduction. It stays attached to its presentation as new feedback arrives; ordinary Help does not replace it. When opened from Recent Feedback, q returns to the same row and column.
Where integrations supply semantic fields, the report groups consecutive voice spans accordingly. Notmuch results, for example, expose Authors, Subject, Date, Message count, and Tags. Headings identify recorded logical voices. Collapsed fields show only headings. Visibility uses configured cues or optional spoken labels; expanded and collapsed are not inserted into heading text.
RET on a heading exposes its text, playback/change controls, voice source, sounds, and limitations. Content uses captured voice annotations. Fields with multiple spans offer Individual voice spans for finer inspection. That heading announces the section without repeating content. A single-span field shows its text and controls once and omits a span count. RET collapses the field again.
Spoken labels such as indent 2 have separate Before speech or After speech headings, distinct from content spans. RET on a label heading, or O within it, replays that label using recorded settings. Edit named voice indent opens the palette voice editor with indent 2 as sample text; it changes every use of that entry in the recorded palette. The report’s recording stays frozen: use the editor’s preview for changes. Labels without a named palette voice remain playable and explain why an editor link is unavailable.
Fields with unapplied drafts announce unsaved change. The report also counts such fields. These indicators update without replacing buttons or return positions; g refreshes draft status, for example after killing a draft buffer. It does not change the recorded or simulated baseline.
The original report layout uses these controls. Outline changes heading and button navigation as described in the next subsection, not playback semantics.
Move between headings; collapsed fields speak the heading, expanded fields also speak annotated content.
Review lines without reading the whole report.
Move between action buttons; RET activates or expands.
Play the original field, span, or label at point, or its labelled simulation.
Play the full original presentation or labelled simulation.
Open or resume a field/span change; on a label, edit its named voice.
Choose a named voice for matching items; on a label, edit its named voice.
Replace, suppress, or restore an exact sound in the field or span.
Preview the full presentation with linked draft changes.
Stop playback.
Refresh draft status.
For example, select Subject, press C, and choose another content voice. In the editor, O plays the original field and P the changed field. Play whole original, also on B, compares with V for the full presentation and linked drafts. Both keep you in the editor. The original remains frozen, including before a change or lifetime is chosen. A simulated report uses Play whole simulation for its frozen baseline.
Return to the report and choose Message count to add a sound After content. It plays once after the complete count, even if the field contains several voices. Subject and count keep separate drafts. Their criteria identify field, integration, mode, and occasion; inspect them before selecting buffer, session, or saved-personal lifetime. Remapping the field does not retune a shared voice.
Choose another voice for matching items opens the compact voice editor. Inspect Applies to and choose How long before saving. Change field opens the full component editor on the same draft. A draft already changing another component stays in the full editor so that work is not hidden.
Sound overrides: replace, suppress, or restore prepares a separate exact-sound rule draft restricted to the field. It is not included in linked-field previews. Restore inherited behavior removes an override made by that workflow in the selected scope. q returns to the report.
Original replay uses the recorded engine and physical voice, including notification voices and fallbacks, with captured tuning even after a palette change. It synthesises speech again at the current reading speed and host pronunciation settings. History stores no audio. Cue and pause order remains, but timing may differ from the first delivery.
Replay requires a completed presentation with complete Omnivox voice receipts. Older, interrupted, truncated, or unsupported records remain inspectable but report replay unavailable rather than guessing a voice. Cues positioned inside speech, or voice changes that cannot be aligned after host text processing, also prevent original replay. An unavailable recorded voice stops playback instead of substituting another. Navigate over the source to capture a new example.
A simulated report labels its baseline throughout the report and change editor. Proposed playback uses current rules, captured buffer context, and linked drafts without applying them. A span can be targeted only when its semantic field, distinct named face, or explicit voice annotation identifies it. Truncated history cannot supply complete playback or a comparable preview. Records missing preview inputs remain readable; capture a new example before previewing.
Each field explains the sources of its voices and sounds, such as a face or personal override, using the contributors actually captured. Unsupported voice families and other limitations appear beside the affected field.
Debug details opens a separate read-only raw snapshot of facts, rule IDs, resolved voices, and backend diagnostics. Opening speaks only a short announcement. q returns to the same report location with fields and drafts retained.
C-e H r t toggles Explain feedback as you navigate for this session. It adds concise semantic explanations after normal presentations, useful for learning an integration’s vocabulary. It is deliberately more verbose than normal use; turn it off when no longer needed.