A local single-track audio analysis framework for factual music maps.
Project description
AudioAtlas
See your track. Keep the judgment yours.
AudioAtlas turns one audio file into a private, portable listening map: key measurements, clear plots, and a short list of places that may be worth checking by ear.
It runs on your computer. There is no account, upload, server, telemetry, or quality score. The result is a folder you can open in any browser and keep with the track.
AudioAtlas provides:
- local analysis with no intentional audio upload;
- portable HTML, Markdown, JSON, and PNG reports;
- Compact, Standard, and Full plot selections over the same complete analysis;
- bounded review prompts that describe measurements without grading the music.
Quick start
Open the live Midnight Studio report to see the complete local-first experience before installing anything. It is generated from the 70.98-second project demo track. The hosted sample contains the static report only; AudioAtlas itself does not upload user audio or generated reports.
The recommended 0.2.0a8 alpha installation is the Python package. AudioAtlas
supports Python 3.11 or newer on macOS, Windows, and Linux where its scientific
Python and audio-decoder dependencies are available.
python -m pip install audioatlas==0.2.0a8
audioatlas --version
audioatlas analyze song.wav --graphs-profile compact
When --out is omitted, the report appears in a folder such as
audioatlas-report-song/ in the current directory. Open report.html inside
that folder. Use --graphs-profile full for all seventeen registered plots.
Release status
Python package and CLI
Available — recommended installation for AudioAtlas 0.2.0a8 alpha. The CLI includes one-file reports, themes, graph profiles, batch reports, manual sections, revision diffs, and song projects.
macOS application
Optional unsigned Apple Silicon technical preview — not signed or notarized. It requires Apple Silicon and macOS 14 or newer, is intended for experienced testers, and is not the recommended installation route. Apple cannot authenticate its developer or notarization status. Do not weaken or disable macOS security controls to run it; use the Python package if normal launch is blocked.
Windows application
Desktop download not included in this alpha — native Windows build and acceptance remain pending. Windows users can use the recommended Python CLI; the native application is neither abandoned nor represented as ready.
Source checkout
uv sync
uv run audioatlas analyze song.wav
The first analysis in a fresh environment may take a little longer while the
scientific libraries initialize. Lightweight commands such as --version,
--help, and themes start without loading the analysis stack.
Try the real demo recordings
The repository includes three intentionally public musical demos: the complete AudioAtlas trailer/demo track, a short solo-guitar recording, and a fuller guitar, koto, cello, and drums arrangement.
# Reproduce the standard Midnight Studio report used by the live site
uv run audioatlas analyze examples/demo_audio/audioatlas_demo.wav \
--out reports/audioatlas-demo \
--graphs-profile standard \
--theme midnight_studio
# Make a clean audio-only input folder, then build an exact three-track catalog
rm -rf reports/demo-audio-input
mkdir -p reports/demo-audio-input
cp examples/demo_audio/*.wav reports/demo-audio-input/
uv run audioatlas batch reports/demo-audio-input \
--out reports/demo-catalog \
--graphs-profile full
# Open the live-demo-style report, catalog, and arrangement report
python -m webbrowser reports/audioatlas-demo/report.html
python -m webbrowser reports/demo-catalog/catalog.html
python -m webbrowser reports/demo-catalog/guitar_koto_cello_drums/report.html
See the recording notes and audio rights notice. These musical demos are not golden test fixtures or threshold-calibration evidence.
Choose how much you want to see
AudioAtlas has one analysis engine. The choices below change report depth and presentation, not the underlying measurements.
| Experience | Command | What changes |
|---|---|---|
| Compact | --graphs-profile compact |
Four essential plots; complete JSON remains available |
| Standard | no extra flag | Fourteen plots and the normal first-read experience |
| Full | --graphs-profile full |
All seventeen registered plots |
Every HTML report opens in Studio and includes a Focus / Studio switch:
- Studio is the polished default, with richer framing, hierarchy, and atmosphere.
- Focus is restrained and information-first.
Both views wrap the same report content and plot pixels. The presentation switch changes framing only.
The polished light theme shown above is the no-flag default. Theme selection styles both the report shell and generated PNG canvases at report-generation time. Midnight Studio is an optional built-in alternative:
audioatlas analyze song.wav --theme midnight_studio
Choose the restrained opening view when generating a report:
audioatlas analyze song.wav --presentation focus
The switch remains available inside the finished report, works offline, and never changes the measured plot content or pixels.
A separate “lite” build is intentionally not maintained. Compact reports use the same trusted analysis engine, which avoids two editions slowly disagreeing about the same track.
Useful recipes
# Pick an output folder
audioatlas analyze song.wav --out reports/song
# Compact first read
audioatlas analyze song.wav --graphs-profile compact
# Restrained opening presentation
audioatlas analyze song.wav --presentation focus
# All plots with a built-in theme
audioatlas analyze song.wav --graphs-profile full --theme midnight_studio
# Analyze one source range
audioatlas analyze song.wav --start 30 --end 62 --out reports/verse
# List themes
audioatlas themes
What you receive
A normal report folder contains:
report.html— the friendly browser report;report.md— a portable text version;summary.json— the complete measurement summary;findings.json— bounded review prompts and their evidence;- PNG plots;
.audioatlas-output.json— a manifest that lets AudioAtlas update its own files without deleting unrelated files.
Local absolute paths are excluded by default, so shared reports do not normally reveal usernames or folder structures.
The HTML report also provides keyboard-accessible plot zoom, direct links between review prompts and their plots, and private Human notes. Notes autosave in local browser storage for that report path and can be copied or exported as text; they are never written into report JSON or sent over a network.
Compare two revisions of the same track
Give related exports the same private revision token when you analyze them:
audioatlas analyze mix-v3.wav --out reports/mix-v3 --track-id "unique-private-token"
audioatlas analyze mix-v4.wav --out reports/mix-v4 --track-id "unique-private-token"
audioatlas diff reports/mix-v3 reports/mix-v4 --out reports/v3-to-v4
The diff reports descriptive B - A changes and which review prompts appeared,
disappeared, or changed. It does not choose a winner. AudioAtlas stores only the
token's SHA-256 digest; matching digests mean the same token was supplied, not
that AudioAtlas recognized the music.
Keep a song's revisions together
For recurring work, create a private local song workspace and add each export in order:
audioatlas project init projects/my-song --name "My Song"
audioatlas project add projects/my-song mix-v1.wav --label "Mix 1"
audioatlas project add projects/my-song mix-v2.wav --label "Mix 2"
Open projects/my-song/project.html. The workspace keeps per-revision reports
and guarded adjacent diffs together. Its YAML configuration records local
source paths for repeatable owner-side use, while generated JSON, Markdown, and
HTML expose only portable filenames and the hashed project identity.
Analyze sections or a folder
Manual sections:
audioatlas sections song.wav --out reports/sections \
--section intro:0:30 \
--section verse:30:62 \
--section ending:62:
Folder catalog:
audioatlas batch /path/to/audio --out reports/catalog
AudioAtlas does not infer song structure. Folder catalogs remain descriptive and do not rank tracks.
What AudioAtlas measures
AudioAtlas currently includes level and loudness context, approximate true peak, clipping and near-clipping counts, RMS and crest timelines, short-term LUFS, stereo correlation, mid/side energy, spectral shape, relative mean band power, onset activity, and chroma pitch-class energy.
Review prompts are checks worth making, not diagnoses. A report may correctly surface no prioritized prompts at all.
What it does not do
- No mix, mastering, loudness, or quality score.
- No automatic EQ, compression, or mastering prescription.
- No cross-track ranking or reference-track winner.
- No genre, instrument, source, key, or automatic section detection.
- No source separation, cloud dashboard, playback engine, or DAW integration.
- No claim that a threshold crossing is audible, bad, or musically wrong.
Alpha status
AudioAtlas 0.2.0a8 is a public alpha. The report pipeline, comparison tools,
and local song-project workflow are usable, but the default review prompts are
still being calibrated on real music. This release adds safer report
replacement, private source binding, GUI failure containment, and stricter
native packaging verification. Signed/notarized or Authenticode-signed desktop
artifacts plus clean-machine acceptance remain future gates; no desktop binary
is generally release-ready. The older .command and .bat launchers remain
convenience wrappers for an already installed CLI, not desktop installers.
Processing time and memory use vary with file length, format, graph profile, and machine. Keep original audio and any reports you care about; alpha feedback about installation clarity, report usefulness, and failures is welcome through the project issue tracker. Reports are descriptive analysis tools, not authoritative mastering judgments.
The temporary Numba compatibility range is documented in Compatibility.
Learn more
- Friendly user guide
- Desktop app and legacy launchers
- Examples
- Finding rules
- Alpha limitations
- Compatibility
- Schemas
- Song-project schema
- Architecture
- Changelog
- Security policy
License
AudioAtlas software and software documentation use the MIT License. The published demo recordings have a separate audio rights notice, including recording-specific CC BY 4.0 terms, an AI-assisted demo-only track, and third-party sound exceptions.
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file audioatlas-0.2.0a8.tar.gz.
File metadata
- Download URL: audioatlas-0.2.0a8.tar.gz
- Upload date:
- Size: 1.4 MB
- Tags: Source
- Uploaded using Trusted Publishing? Yes
- Uploaded via: twine/6.1.0 CPython/3.13.13
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
9ccb7566ac5c8cf4c5eaaaa8728b99a80f7f4b3da283350901b9754a83733c0c
|
|
| MD5 |
9241609c4e9e3e57dd5d6aa4d967b877
|
|
| BLAKE2b-256 |
114cf5e458d4b9e6432b18d0a0850930d8e0cf7e0350a4624c274ffae8f19c68
|
Provenance
The following attestation bundles were made for audioatlas-0.2.0a8.tar.gz:
Publisher:
release.yml on CharlesMish/AudioAtlas
-
Statement:
-
Statement type:
https://in-toto.io/Statement/v1 -
Predicate type:
https://docs.pypi.org/attestations/publish/v1 -
Subject name:
audioatlas-0.2.0a8.tar.gz -
Subject digest:
9ccb7566ac5c8cf4c5eaaaa8728b99a80f7f4b3da283350901b9754a83733c0c - Sigstore transparency entry: 2329573820
- Sigstore integration time:
-
Permalink:
CharlesMish/AudioAtlas@f04a6f854a9c6a0bf308addfa34aa676f1120fc5 -
Branch / Tag:
refs/heads/main - Owner: https://github.com/CharlesMish
-
Access:
public
-
Token Issuer:
https://token.actions.githubusercontent.com -
Runner Environment:
github-hosted -
Publication workflow:
release.yml@f04a6f854a9c6a0bf308addfa34aa676f1120fc5 -
Trigger Event:
workflow_dispatch
-
Statement type:
File details
Details for the file audioatlas-0.2.0a8-py3-none-any.whl.
File metadata
- Download URL: audioatlas-0.2.0a8-py3-none-any.whl
- Upload date:
- Size: 166.0 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? Yes
- Uploaded via: twine/6.1.0 CPython/3.13.13
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
f4546ca18499299a8108312a95f9e1d527565ecc8e8fea62f64144b7b5a7fde7
|
|
| MD5 |
f275a62fb8048c685b8a26508df04524
|
|
| BLAKE2b-256 |
7e0d6fe5f1d807efddb2a9f84af0e470454efbf28ea72b2378286da704268200
|
Provenance
The following attestation bundles were made for audioatlas-0.2.0a8-py3-none-any.whl:
Publisher:
release.yml on CharlesMish/AudioAtlas
-
Statement:
-
Statement type:
https://in-toto.io/Statement/v1 -
Predicate type:
https://docs.pypi.org/attestations/publish/v1 -
Subject name:
audioatlas-0.2.0a8-py3-none-any.whl -
Subject digest:
f4546ca18499299a8108312a95f9e1d527565ecc8e8fea62f64144b7b5a7fde7 - Sigstore transparency entry: 2329573911
- Sigstore integration time:
-
Permalink:
CharlesMish/AudioAtlas@f04a6f854a9c6a0bf308addfa34aa676f1120fc5 -
Branch / Tag:
refs/heads/main - Owner: https://github.com/CharlesMish
-
Access:
public
-
Token Issuer:
https://token.actions.githubusercontent.com -
Runner Environment:
github-hosted -
Publication workflow:
release.yml@f04a6f854a9c6a0bf308addfa34aa676f1120fc5 -
Trigger Event:
workflow_dispatch
-
Statement type: