[P1] Complete accessible onboarding, explanations, and recovery states #15

Parent plan: #1
Priority / gate: P1 / M3
Requirement: F11 in the PRD
Design: Technical design
Depends on: #3, #12, #13, #14, #18, #19, #20

Outcome

Make the core sketch → function → sound → export path understandable and usable without relying only on freehand input or visual plots.

Scope

• Include soundboard Save/name, focused-pad Enter/Space playback, Stop, rename/reorder/remove/undo, and Open in editor in the keyboard and status-announcement review.
• Explain waveform time/amplitude axes versus spectrum frequency/component-level axes, with separate time/gain and time/pitch-Hz envelope lanes. Demonstrate amplitude scaling versus DC translation without equating signed sample height with volume.
• Add concise first-use instructions and in-context explanations for phase versus time, waveform versus pitch/gain envelopes, Hz versus linear gain, source versus rendered audio, and sub-audible timelines.
• Complete keyboard navigation, visible focus, shortcuts, numeric editing, accessible equations/metrics, and status announcements.
• Review non-color curve distinctions, reduced-motion behavior, control labeling, and error recovery.
• Exercise invalid-import, unsupported-audio, renderer, storage, and resource-limit states.
• Include local file selection, numeric trim/offset, carrier/modulator selection, mode/depth/attack/release/loop controls, source audition, bypass/A–B, and asset-bearing save/reopen in the accessibility review. Explain volume follow versus signed ring multiplication and oscillator pitch versus clip varispeed.

Acceptance criteria

• A keyboard user can save a named sound, replay it, stop it, manage it, and reopen an editable copy; loading/playing/error states are distinguishable without relying on color.
• First-use guidance correctly explains that cycle spacing within a fixed time window expresses frequency, waveform height around zero expresses amplitude, and moving the whole curve upward adds DC offset.
• A keyboard user can create/edit a signal numerically, select its representation, play/stop, and export without a pointer.
• Focus remains sensible after tool changes, dialogs, errors, and imports; status messages do not overwhelm assistive technology.
• Every canvas-only fact essential to the workflow has a text/numeric counterpart.
• The user can identify whether a displayed equation describes the source, an approximation, or rendered audio.
• Failure states preserve recoverable work and provide a clear next action.
• In a five-person formative test, at least four users draw, hear, and find the function within two minutes without coaching, or resulting fixes are completed and retested.
• Users can identify each oscillator without relying on color alone, edit its pitch and gain independently, and distinguish plot visibility from mute/solo.
• A keyboard user can choose an uploaded or saved modulation source, set target/effect/timing, preview/bypass, export, and save/reopen it; numeric equivalents, varispeed timing implications, source/result identity, decode state, and recoverable errors are understandable without relying on color.

Validation

Manual keyboard and screen-reader sessions, automated accessibility checks where useful, and recorded formative-usability observations with no uploaded signal content required.

Boundaries

A formal accessibility certification or a dedicated mobile-first interface.

Implemented and verified

The application has labeled numeric editing, native keyboard controls, visible focus, non-color curve cues, source/result explanations, explicit modulator audition, local save/recovery states, and error-preserving imports/exports. Automated WCAG-tagged axe checks and narrow-screen layout checks pass. PR #30 adds the in-app quick guide, arrow-key workspace selection, playback position, and the complete user guide.

Validation CI · Validation report · implementation PRs #21–#30.

Remaining acceptance — keep open

• Manual keyboard and screen-reader review, including focus behavior and equivalents for essential canvas information.
• The five-person formative test and any fixes it reveals.

The automated accessibility result is not represented as human usability evidence.