Repository navigation
Releases
SONOLYTICA 0.2.4 — human editing layer
SONOLYTICA 0.2.4
The human editing layer: interactive reports and subtitle/notes export.
This is a presentation release. No schema change, no model change, no
benchmark rewrite. Agents still use the same document, query and MCP tools.
Report
- Transcript search with a live match count.
- Click a range to copy it; with
--media, the report embeds a local player and
the same click seeks to that moment. sonolytica report talk.json --out talk.html --media talk.mp4- The report remains one self-contained HTML file plus copied keyframes. Inline
JavaScript only; no network requests.
Export
New command over the same document:
sonolytica export talk.json --format srt --out talk.srt
sonolytica export talk.json --format vtt --out talk.vtt
sonolytica export talk.json --format markdown --speakers --ocr --out talk.md- SRT/WebVTT ranges are millisecond-exact document ranges.
--speakersprefixes cues with the diarized speaker when available; it does
not invent a speaker if diarization did not run.- Markdown can include OCR text from
--ocr. - No text is rewritten or summarised.
Two front doors, still
- Human: analyze → report → search/seek, or export subtitles/notes.
- Agent:
sonolytica mcp→ cited answers; unchanged 14 tools and skill. - Both use one document and one cache;
docs/START-HERE.mdcovers both.
Tests
187 model-free / 204 all-features.
Release packaging is unchanged: Windows 10/11 x64, macOS arm64, Linux
x86_64/arm64, Docker GHCR, checksummed installers.
SONOLYTICA 0.2.3 — two front doors
SONOLYTICA 0.2.3
Two front doors, one document: a human report path and an AI-agent MCP path.
This is an onboarding and skill release. No schema change, no model change, no
benchmark rewrite.
Human path
docs/START-HERE.mdwalks from install to an HTML report
with a public-domain 11-second sample.analyze→reportproduces one shareable page with chapters, quotes,
keyframes and OCR;scripts/demo.shwraps the same path.- No MCP client, JSON reading or query syntax required.
Agent path
- The same guide gives the MCP config for macOS/Linux and Windows, a first
prompt, the expected tool sequence (analyze_media→query_brief→
query_ask→query_search/query_words→query_verify) and a smoke test. - The skill now chooses the right mode explicitly:
- human-facing request → analyze + report;
- agent-facing question → query/MCP with citations;
- both → analyze once, reuse the document.
sonolytica setupinstalls the updated skill; re-run it after upgrading.
Unchanged
- Schema 1.6.0, 14 pinned models, 17 query subcommands, 14 MCP tools.
- 184 model-free / 201 all-features tests; the same benchmark and accuracy
claims. - Windows 10/11, macOS arm64 and Linux x86_64/arm64 release packaging.
Install
# macOS / Linux
curl -fsSL https://raw.githubusercontent.com/neksand/sonolytica/main/install.sh | sh
sonolytica setup
# Windows 10/11 x64 (PowerShell)
irm https://raw.githubusercontent.com/neksand/sonolytica/main/install.ps1 | iex
sonolytica setupThen choose a path in docs/START-HERE.md.
SONOLYTICA 0.2.2 — real-media trust fixes
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 arenull/empty when
there is no evidence; loudness remains valid. Clean speech (the 11-second
JFK fixture) reportsmusic not detected; a click track and a sustained
triad still pass. Query JSON exposespresent: 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 reportscore 0.000while still publishing a BPM. - Fallback warning noise. A missing optional fallback (pyannote when
Sortformer is present) no longer emitsprovider_unavailableon every run.
If no provider covers a requested capability, one warning still fires;
sonolytica doctorstill lists every unavailable model. - Evaluation CLI.
sonolytica-eval report DOCUMENTnow works positionally;
--documentremains a deprecated alias.weraccepts
--reference-text/--hypothesis-textfor raw strings and-for stdin,
not only file paths. - Language-aware ASR choice. English uses
whisper-base.enby default;
non-English requests preferwhisper-smallwhen installed; an explicit
--model(includingauto) wins.sonolytica doctorprints 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-smallis not the English default; use--model whisper-small.- Real overlapping speech still needs an overlap-aware diarizer.
SONOLYTICA 0.2.1 — Windows 10/11 x64
SONOLYTICA 0.2.1
Windows 10/11 x64 support, with the same local, citable document as macOS and
Linux.
This is a platform release: no schema change, no benchmark regression, no new
model. The pipeline, 14 pinned models, 17 query subcommands, 14 MCP tools, cache
and HTML report are the same code.
Windows
- Native x86_64-pc-windows-msvc binary, one zip for Windows 10 and 11.
Both share the Win32 x64 ABI; separate binaries would be byte-identical. install.ps1: downloads the release zip, verifies SHA-256, installs
sonolytica.exeandsonolytica-eval.exeunder
%LOCALAPPDATA%\sonolytica\bin.%USERPROFILE%fallback for the model/cache/skill directories; native
PowerShell does not setHOME.- Windows font fallback for the OCR drawing test.
- CI builds and tests the model-free workspace on Windows and builds the full
provider stack (ort, whisper.cpp, Kaldi fbank) in release mode. - FFmpeg remains an external runtime requirement:
winget install Gyan.FFmpeg.
Windows 8/8.1, Windows 7, Windows on ARM and 32-bit Windows are not
supported. Windows 8 is end-of-life and outside the current Rust/MSVC, ONNX
Runtime and FFmpeg toolchains; a separate legacy build would be an unmaintained
second product. Use a supported OS, Linux, or Docker.
Install
Windows PowerShell:
irm https://raw.githubusercontent.com/neksand/sonolytica/main/install.ps1 | iex
sonolytica setupmacOS arm64 / Linux x86_64 / Linux arm64:
curl -fsSL https://raw.githubusercontent.com/neksand/sonolytica/main/install.sh | sh
sonolytica setupPlatforms
| Platform | Artifact |
|---|---|
| Windows 10/11 x64 | sonolytica-x86_64-pc-windows-msvc.zip |
| macOS arm64 | sonolytica-aarch64-apple-darwin.tar.gz |
| Linux x86_64 | sonolytica-x86_64-unknown-linux-gnu.tar.gz |
| Linux arm64 | sonolytica-aarch64-unknown-linux-gnu.tar.gz |
Every archive has a matching .sha256.
Unchanged evidence
| Claim | Value |
|---|---|
| hard 4-minute fixture | 16/16 structured; blind judge 15/16, 19 valid citations |
| long 20-minute meeting | 16/16*; 4/4 speakers, DER 6.40 % |
| easy fixture | 8/8 |
| 20-minute full audio pipeline | 45.1 s (26× realtime), 2.07 GB peak RSS |
| OCR | 12/12 slide lines; 7/7 CJK terminal |
| tests | 173 model-free / 188 all-features |
* Q4 corrected: the fvhrk RTTM has only spk02 in the questioned window;
the old overlap spans were merge artifacts. See docs/bench/overlap.md.
Known limits
- Real overlapping speech needs an overlap-aware diarizer mask; Parakeet
recovered the hard-fixture line only with a reference mask. - Sortformer caps at 4 speakers.
- Captions are keyframe observations, not free-form visual Q&A.
- Whisper word alignment is English-only.
- Music has tempo/key/sections/loudness, no notes or stems.
These are the same boundaries as 0.2.0 and remain on the roadmap in the README.
License
Code MIT OR Apache-2.0. Models keep their own permissive licenses and are never
vendored. See LICENSE-MIT, LICENSE-APACHE, NOTICE and docs/MODELS.md.
SONOLYTICA 0.2.0
SONOLYTICA 0.2.0
Media becomes a document an agent can cite. Local, typed, timestamped, with
provenance on every claim.
Highlights
- AOSC streaming diarization: 4/4 speakers on the 20-minute
fvhrk, DER
6.40 % (was 2/4, 15.99 %); no regression on 1–4 speaker files. - Florence-2 keyframe captions (
--caption, MIT, 550 MB): descriptions
such as suit/tie/arms/flags/smile; model-generated observations, not OCR. - MoveNet body pose (
--pose, Apache-2.0): 17 COCO keypoints per person per
keyframe, detection only. - Analysis cache: 0.01 s hit vs 7.64 s first run; artifacts travel with the
document,analysis_cachedis explicit. query ask: one deterministic evidence pack with absent terms, instead
of several searches.- MCP stdio server (
sonolytica mcp): 14 tools for Claude Desktop, Cursor
and IDE agents; the same binary and cache, no duplicate logic. - One-command onboarding:
sonolytica setup(core models + skill),
--fullfor vision;install.shpoints at it. - Docker: 790 MB image, FFmpeg 7, models in a
/datavolume; verified
end-to-end on arm64. - Schema 1.6.0 (pose + captions), 14 pinned models, 173 model-free / 188
all-features tests.
Measured
| Fixture | Score | Cost |
|---|---|---|
| hard 4 min | 16/16 | paid median 14 773 fresh; free 19 396, $0 |
| easy 14.4 s | 8/8 | paid median 8 818 fresh; free 18 224, $0 |
| long 20 min | 16/16* | paid median 13 443 fresh; free 31 570, $0 |
| visual boundary | 16/16 refusal; 16/16 caption-aware | 17 652 / 31 961 free |
* Q4 corrected: the fvhrk RTTM has only spk02 in the questioned window;
the old overlap spans were merge artifacts. See docs/bench/overlap.md.
Install
curl -fsSL https://raw.githubusercontent.com/neksand/sonolytica/main/install.sh | sh
sonolytica setup # core models + skill
sonolytica setup --full # OCR, faces, objects, pose, Florence
sonolytica mcp # MCP clientsKnown limits
- Open visual semantics are partial: captions/detectors, no free-form VQA.
- Sortformer caps at 4 speakers; a full four-slot hypothesis is a lower bound.
- Real overlap needs an overlap-aware mask; Parakeet recovered the hard-fixture
line with a reference mask, but the AOSC S4 activity there is ≤0.011. - Non-English word alignment is not implemented; Whisper token times remain.
- Music has tempo/key/sections/loudness, no notes or stems.
- A very short clip can be cheaper with a bare Whisper call.
License
Code MIT OR Apache-2.0. Models keep their own permissive licenses and are not
vendored; see docs/MODELS.md and NOTICE. FFmpeg is a subprocess.