Default voice and language-phonemizer selectors for new projects and score imports.
Piano-roll note editing actions: copy selected notes, preview and place pasted notes, move notes by semitone or octave, and undo the resulting edits. Paste uses a stable anchor, avoids collisions where possible, and reports notes that do not fit. The actions are available from the Edit menu and can be assigned keybindings.
Localized application interface translations, including setup, import, settings, editor controls, voice controls, and progress messages. Missing translations fall back to the default language.
A language phonemizer picker in score import. Choose Automatic or a specific installed language package; the choice is stored on the imported voices and can be changed later.
The bundled Japanese portable CV/VCV/CVVC language package `lang.ja.contrib.ucv`.
A detailed language-plugin contract reference for package authors, including the input track shape, notes, gaps, phoneme plans, timing edits, diagnostics, provenance, and supported engine phonemes.
A canonical `Time` API for plugin authors. `Time` values are numbers measured in 100 ns units, with conversion helpers for seconds, milliseconds, microseconds, and nanoseconds. The utility is available as `Time`, `time`, and `globalThis.Time`.
Changed
The existing first-run setup now uses the saved default voice and phonemizer selections and is available through the translated setup flow.
The voice mixer presents clearer per-voice controls and status: readable cast names, level meters, gain and pan controls, mute/solo controls, and language-package selection.
Render progress identifies the voice or cast being processed and reports an estimated time remaining without interrupting project work.
Playback applies voice lookup, mute/solo state, gain, and meter routing consistently before audio reaches the output callback.
Plugin authors can use the built-in TypeScript language-contract types without maintaining a separate domain declaration. Structured pitch expression, vibrato, metadata, JSON extensions, and module provenance are available to consumers.
Plugin timing fields prefer the shorter names `start`, `end`, and `boundaryOffset`; older `start100ns`, `end100ns`, and `boundaryOffset100ns` package data remains accepted.
Language-plugin compatibility is explicit: SDK 1.2.1 packages require the monotonic schema contract and an Amadeus host version of beta3 or newer. Incompatible packages are rejected when registered.
Fixed
Deleting a voice now preserves the active editing voice and keeps project, reference-audio, and progress state synchronized.
0.1.0-beta.2
Added
Additional acceleration choices for supported systems, including AMD and NVIDIA options.
A more practical debug panel with collapsible sections and scrolling.
Changed
Refactored the UI routine for performance-critical operations.
Improved settings selection layout and made the available choices easier to scan.
Improved the workflow for selecting acceleration families and runtimes.
Improved plugin and language selection updates after a plugin is enabled, disabled, or refreshed.
Fixed
Note and timing controls are less likely to steal clicks from a neighboring control.
Timing-boundary dragging stays within the valid range of neighboring phones and notes.
Zooming and panning keep visible editing boundaries aligned with the score.
Audio previews no longer select cache data from an unrelated project or track.
Save As and voice-stem export preserve project references.
Duplicate voice-stem outputs are detected before files are written.
Removed
The unused gap rail from the Timing editor.
0.1.0-beta.1
Available
The beloved editor
Voice render engine
MIDI and MusicXML score import
Project containers with rendered voice
Vietnamese pronunciation and lexical-tone plugin
Built in pitch processors
JavaScript plugin
Linux and Windows build
ORT CPU, CUDA, OPENVINO
Limits
MusicXML primarily targets MuseScore and Dolet for Sibelius
Lyric unrolling, importer and exporter registration, and note-scoped parameters are still limited
See the manual for current user behavior and extension contracts.
Roadmap
The roadmap is split between near-term work on the core workflow and longer-term additions to the editor and plugin ecosystem.
WIP
Improve controls over the voice and model pick.
Improve the flexibility and ease of language plugin development.
Add an interchangeable VOICEVOX backend, a system currently limited to Japanese.
