Read the Editor/helper fingerprint, host render lease, and verified voice cache without launching VOCALOID.

Verify one local voice with a bounded temporary render in the original Editor; cache the result by voice and host fingerprint.

Browse phrase choices, validate a blueprint, or analyze diversity; never generates melody.

Align Japanese morae to stable rhythm slots as a manifest-compatible phrase, with explicit melisma and small-っ timing decisions.

Measure WAV loudness, dynamics, peaks, clipping, and stereo balance without modifying the file.

Import a V3/V4 VSQX vocal score into a new stable-ID manifest and record the source with warnings for unsupported data.

Apply one strict action payload from vocaloid://edit-song-contract; dry_run returns hashes, music-aware diff, theory/intent deltas, and preflight without writing.

Append/resume, query, report, or publish decisions.

Create the revisioned canonical song manifest; derive VSQX/WAV artifacts later through the build workflow.

Verify and summarize a canonical song manifest, including its revision, stable IDs, inventories, intent, and artifact provenance.

Run non-mutating structure, theory, vocal, arrangement, mix, artifact, intent, and host checks. Optional repair hypotheses stay deterministic, theory-only, revision-bound, and require vocaloid_edit_song dry_run plus listening.

Render a current VSQX through the original Editor, verify the WAV, and atomically record a provenance-bound master artifact.

Two-pass loudness-master a current mix/master, report deterministic peak/LRA transparency risk for source-vs-delivery A/B, and atomically record verified provenance.

List, music-diff, or monotonically restore verified immutable song-manifest history snapshots.

Render section or 20–30s anchor previews and single-question loudness-matched A/B evidence; this tool never chooses a winner.

List or persist qualitative listening notes, including hash-bound keep/reject/revise evidence for phrase intent; metrics cannot self-approve.

Fetch one semantic section and only the vocal, harmony, and arrangement objects that intersect it.

Build a verified accompaniment WAV and native VSQX from one immutable manifest revision, then record both artifacts atomically.

Render selected arrangement tracks as isolated verified WAV stems with explicit pan/effect policy and manifest provenance.

Render selected vocal tracks individually through the original Editor as verified stems with explicit pan/send policy and provenance.

Compare stable vocal events with a verified isolated stem and record reproducible pitch, timing, energy, diction, and vibrato evidence.

Render selected verified stems through manifest automation/effects without hidden normalization, then record the measured mix and filter graph.