Skip to content

SONOLYTICA 0.2.2 — real-media trust fixes

Choose a tag to compare

@github-actions github-actions released this 13 Nov 10:41
· 3 commits to main since this release

SONOLYTICA 0.2.2

Trust fixes from real-media dogfood: no ghost music, no fallback noise,
clearer evaluation commands, language-aware ASR choice.

No schema change, no new model, no benchmark rewrite.

Fixed

  • Ghost music on speech. The DSP now gates musical evidence before the
    document is built. BPM, key, chords, beats and sections are null/empty when
    there is no evidence; loudness remains valid. Clean speech (the 11-second
    JFK fixture) reports music not detected; a click track and a sustained
    triad still pass. Query JSON exposes present: false/true.
  • Meaningful tempo confidence. Tempo confidence is now a normalized
    cross-correlation, comparable across sparse clicks and dense speech. The old
    raw ratio could report score 0.000 while still publishing a BPM.
  • Fallback warning noise. A missing optional fallback (pyannote when
    Sortformer is present) no longer emits provider_unavailable on every run.
    If no provider covers a requested capability, one warning still fires;
    sonolytica doctor still lists every unavailable model.
  • Evaluation CLI. sonolytica-eval report DOCUMENT now works positionally;
    --document remains a deprecated alias. wer accepts
    --reference-text/--hypothesis-text for raw strings and - for stdin,
    not only file paths.
  • Language-aware ASR choice. English uses whisper-base.en by default;
    non-English requests prefer whisper-small when installed; an explicit
    --model (including auto) wins. sonolytica doctor prints the default.

Default model

whisper-base.en remains the fast English default. whisper-small is the
quality option and the automatic choice for explicit non-English languages.
This avoids silently changing the 26× realtime benchmark; switching the English
default requires a separate measured decision.

Upgrade

Re-run sonolytica analyze on documents produced by 0.2.1 to apply the music
gate. Existing documents are still readable: query treats low-confidence music
as present: false and hides BPM/key/chords.

Unchanged evidence

The hard/long benchmark, DER 6.40 %, 14 models, schema 1.6.0, MCP tools and
Windows/macOS/Linux packaging are unchanged. Tests: 184 model-free / 201
all-features
. The honest ASR number remains
6.05 % WER / 2.25 % CER on LibriSpeech test-clean; 0.00 % on an 11-second
JFK clip is a sanity check, not a general accuracy claim.

Known limits

  • Beatless music with weak or noisy triads may be under-reported by the
    conservative gate. It will not hallucinate music on speech.
  • whisper-small is not the English default; use --model whisper-small.
  • Real overlapping speech still needs an overlap-aware diarizer.