SpeakTab 10.14.31 — saved songs, teacher setup and clearer piano reading

SpeakTab 10.14.31 adds saved song workspaces, clearer piano labels, teacher setup and onboarding, practice drums, and contracted textual braille export.

Download and begin

Download SpeakTab-Windows.zip, extract the entire ZIP, keep its recognition runtime, licenses and guides together, and run SpeakTab.exe. Keep your original music and a previous app copy. Developers should download SpeakTab-Source.zip for the complete handoff; GitHub's generated source archives are not that package.

The 95-page User Manual and 5-page Quick Start are available below as Word documents and plain text. The full manual also has an HTML edition. Instructions cover all 154 current menu actions, with generic examples, menu access keys, first-use procedures, expected results and recovery steps. No prior SpeakTab experience is assumed.

Save and return to music

Opening music no longer creates automatic reading copies. Ctrl+S saves or updates one arrangement under Saved Readings > Songs. Closing or switching away from new or changed work asks whether to save, leave without saving or cancel. File > My readings offers the previous lesson summary, Resume and Start over. Separate source files or parts can represent different arrangements.

Keep a separate chord or lyric sheet beside a detailed score through File > Companion chord or lyric sheet. Save the workspace to retain the imported companion. A chord chart does not supply missing picking notes. Export reading copy as text or HTML makes a separate outside copy.

One teacher for guitar and piano

Voice > AI teacher setup explains naming the teacher, separate instrument preferences, voice connection and onboarding. Start AI teacher lesson explicitly starts the configured live teacher. Spoken onboarding saves the profile and requests questions immediately; missing setup opens a focused explanation with a connection-settings button. An active teacher receives the request directly, and a connecting teacher retains it without duplicate first questions. F9 stops the lesson. Live teaching requires a voice API key and separately billed API access; the free reader microphone remains separate.

Saved workspaces can hold factual lesson summaries. Ask to review a saved arrangement, start it over, switch to another saved arrangement or suggest an exercise for a passage. Read the actual save/open result; AI responses alone do not prove the action succeeded.

Piano, rhythm and textual braille

• Piano labels use the current supplied treble or bass clef, with staff numbers retained for navigation. A single voice needs no voice-number announcement. Multiple voices are numbered locally within each staff, preserving source identities. Missing clefs stay unspecified; staff labels do not determine hands.
• Practice > Practice drums — no song needed offers four original synthesized patterns with tempo and volume controls. They work independently of a song and are not synchronized backing for score playback. The existing loop window contains optional gradual-speed settings; its duplicate menu entry was removed.
• File > Export textual braille BRF defaults to contracted English UEB prose through bundled Liblouis. It exports text descriptions, not braille music. It does not add MusicXML export or operate an embosser. Matching Liblouis source and licenses are included.

Verification and limits

The root release gate and all 57 focused software checks passed. Both complete packages passed executable/source/document, manifest, runtime, license and privacy checks. Current and previous public updaters extracted the actual final package with the expected executable hash. Eden reports the practice drums and live teacher work; the initially silent teacher was traced to the microphone being off. This is not exhaustive certification of every device, onboarding conversation or AI response.

PDF/OCR and AI recovery remain experimental. Compare notes, rhythm, lyrics and coverage with originals. Broader JAWS/NVDA/Braille, microphone, audio, musical accuracy and live-model checks remain ongoing. Playback remains sampled guitar sound. Private music, saved readings, profiles, credentials and private work logs are excluded from the distribution packages.