Strata

A self-hosted ambient sound mixer for layering music, soundscapes, and audio — runs in the browser, lives in your system tray.

---

What it does

Strata lets you build a personal ambient audio environment. Load looping bed tracks (rain, music, ambience) at independent volumes, queue up longer pieces to play in sequence, and save the whole state as a preset you can recall instantly.

Audio comes from local files or any URL yt-dlp supports. Chapters and tracklist timestamps are parsed automatically into segments you can navigate track by track.

---

Features

Mixer

• Beds zone — multiple tracks loop simultaneously, each with independent volume
• Queue zone — sequential playback with drag-to-reorder, loop toggle, and master volume
• Segment navigation — ‹ / › buttons on each bed track jump between segments; the active chip highlights live as the track plays
• Transport controls — global ⏮ / ▶ / ⏭ with Spotify-style prev (restarts track if past 3 s, goes to previous otherwise)
• Presets — save and restore full mixer states; mark one as the startup default

Library

• Browse — filter by All / By Channel / By Mood; sort by date, title, or play count
• Search — live filter across titles and segment names; matching cards expand their segment list automatically
• Row expansion — expanding segments on one card opens the whole grid row together
• Stats — listening time (all time / 30 days / 7 days), most played, recently played

Track editor

• Download from URL — paste any URL; yt-dlp fetches audio + thumbnail in the background
• Auto-segmentation — chapters from yt-dlp and tracklist timestamps from descriptions are parsed automatically
• Manual segments — add, edit, reorder; paste a tracklist block and click Parse to auto-generate
• Custom thumbnails — upload your own image per track
• Mood tags — comma-separated with autocomplete from your corpus

---

Installation

Desktop app — Windows · macOS · Linux

Download the installer for your platform from the Releases page:

Platform	File
Windows	Strata-Setup-x.x.x.exe — wizard installer, installs per-user, no admin required
macOS	Strata.dmg — drag to Applications; right-click → Open on first launch (Gatekeeper)
Linux	Strata-linux-x86_64.tar.gz — extract and run ./Strata/Strata

After launch, Strata lives in the system tray (Windows/Linux) or menu bar (macOS). Click Open Strata to open the UI in your browser.

User data (tracks, presets, database) is stored in:

• Windows: %APPDATA%\Strata
• macOS: ~/Library/Application Support/Strata
• Linux: ~/.local/share/Strata

---

Self-hosted (Docker)

The docker-compose.yml is written for a homelab running Traefik as a reverse proxy. Adjust the Host labels and volume path to match your setup.

docker compose up -d --build
docker compose logs -f strata

The container expects a TZ environment variable (e.g. in a .env file):

TZ=Europe/Rome

The media volume must be writable by the container. On the reference deployment it maps /mnt/raid/strata/media → /media.

---

Run from source

# 1. Install dependencies
pip install -r strata/requirements.txt   # skip gunicorn on Windows/macOS
pip install -r desktop/requirements.txt

# 2. Generate placeholder icons (once)
python desktop/assets/generate_icon.py

# 3. Run
python desktop/main.py

Or without the tray, using the Flask dev server:

cd strata
MEDIA_ROOT=./media flask run   # Windows: set MEDIA_ROOT=.\media

See desktop/README.md for building a distributable installer from source.

---

Stack

Layer	Technology
Backend	Python 3.12, Flask 3, Waitress (desktop) / Gunicorn (Docker)
Audio download	yt-dlp + ffmpeg
Frontend	Vanilla JS, Web Audio API
Storage	Flat files (JSON + mp3 + thumbnail) + SQLite (telemetry)
Desktop wrapper	pystray, PyInstaller
Container	Docker, linux/arm64

Project structure

strata/                 # Flask app (shared by both targets)
├── app.py              # API + download jobs + DB telemetry
├── wsgi.py             # Gunicorn entry point (Docker)
├── requirements.txt
├── Dockerfile
├── templates/index.html
└── static/
    ├── css/style.css
    └── js/app.js
desktop/                # Native desktop wrapper
├── main.py             # pystray entry point, starts waitress
├── requirements.txt
├── README.md           # Desktop build guide
├── assets/
│   ├── generate_icon.py
│   └── bin/            # Place yt-dlp + ffmpeg binaries here
└── build/
    ├── strata.spec     # PyInstaller spec
    └── installer.iss   # Inno Setup (Windows)
docker-compose.yml

Media layout:

<MEDIA_ROOT>/
├── tracks/<uuid>/
│   ├── audio.mp3
│   ├── thumbnail.jpg
│   └── meta.json
├── presets/<name>.json
└── telemetry.db

---

Environment variables

Variable	Default	Description
MEDIA_ROOT	/media	Root path for tracks, presets, and telemetry DB
TZ	—	Timezone for timestamp display
FLASK_ENV	production	Set to development for debug mode

API routes

Method	Path	Description
GET	/api/tracks	List all tracks
POST	/api/tracks	Upload an audio file
PATCH	/api/tracks/<id>	Update track metadata
DELETE	/api/tracks/<id>	Delete track and all its files
GET	/api/tracks/<id>/thumbnail	Serve thumbnail
GET	/api/tracks/<id>/audio	Serve audio
POST	/api/download	Queue a yt-dlp download job
GET	/api/jobs/<id>	Poll download job status
GET	/api/presets	List presets
POST	/api/presets	Save a preset
DELETE	/api/presets/<name>	Delete a preset
GET	/api/stats	Listening time and play-count stats
POST	/api/telemetry	Record a play event

Segment auto-detection

When a track is downloaded, segments are built from (in order of preference):

1. yt-dlp chapters — from the video's native chapter markers
2. Description parsing — timestamps found in the description; supports [00:00], 00:00, 01:01:30, with or without track numbers or trailing URLs

If description parsing yields more segments than native chapters, it wins. You can also paste a tracklist manually in the editor and click Parse.

---

Disclaimer

Strata is a personal-use tool. You are responsible for ensuring that any audio you download complies with the terms of service of the source platform and applicable copyright law. The authors make no representations regarding the legality of downloading specific content.

---

License

MIT © Claudio Bendini