lucidadl 0.1.1

Fast, parallel CLI music downloader for lucida.to (Qobuz, Amazon Music) with local transcoding and playlist import.

lucidadl

Fast, parallel command-line music downloader built on top of lucida.to. Search or paste a URL, download tracks and albums in parallel over plain HTTP, organize them by tags, transcode to the format/bitrate you want, and import playlists.

A vibe-coded project. lucidadl was built quickly and AI-assisted ("vibe coding"). It wraps and automates downloading from lucida.to — inspired by two existing open-source lucida downloaders we looked at (see Credits) and adding the features I wanted on top: parallel downloads, local transcoding, tag-based organization, playlist import, existence-aware dedup, and an interactive menu.

Disclaimer. This is a personal-use tool, similar in spirit to yt-dlp. You are responsible for complying with the terms of service of lucida.to and of the source services, and with the copyright law of your jurisdiction. The authors are not affiliated with lucida.to or any streaming service. Use it for content you are entitled to download.

Features

• Search or URL — track "artist - title", album "artist - album", or paste a Qobuz/Amazon URL.
• Parallel downloads over HTTP (--jobs N) — no browser kept open, low RAM.
• Albums = track by track, expanded and downloaded in parallel (faster than the native album zip).
• Local transcoding with ffmpeg (--to mp3 --bitrate 320k) — bundled, nothing to install. Tags and cover art preserved.
• Tag-based organization into Artists/<Artist>/<Album>/… (falls back to the source's artist/album metadata when a file has no embedded tags, so nothing lands in "Unknown"); playlists go under Playlists/<name>/, kept separate from artists.
• Watchlists with dedup (tracks / albums read a file, skip what's done — but a file you deleted is re-downloaded; --force ignores the memory entirely).
• Playlist import — paste a playlist link and get every track via lucida. Apple Music is wired up today; adding more sources is easy (PRs welcome).
• Interactive search, service fallback (Qobuz → Amazon), retry of failures.
• Interactive menu (lucida ui, or just lucida) and live progress bars — one bar per parallel download, in any real terminal.

How it works (Cloudflare)

lucida.to is behind Cloudflare and plain HTTP gets a 403. lucidadl solves the challenge once in a real browser (setup), caches the cf_clearance cookie, then runs everything else over httpx — no browser stays open. The browser is only re-opened briefly if the cookie expires. Solving the challenge needs a logged-in desktop session (it can't run on a locked/headless server).

Requirements

• Python 3.10+
• A desktop session for the one-time Cloudflare setup (Windows/macOS/Linux).

Install

pip install adds a global lucida command (alias: lucidadl).

Already using lucida-downloader? Its binary is also called lucida, so the two clash on your PATH. Just use the lucidadl alias for this tool (e.g. lucidadl track "…", lucidadl ui) — every command below works the same with lucidadl in place of lucida.

Recommended — isolated, on PATH (pipx):

pip install --user pipx && python -m pipx ensurepath   # once, if you don't have pipx
git clone https://github.com/Jude-A/lucidadl
pipx install ./lucidadl
pipx run playwright install chromium                    # one-time: download the browser

Or with plain pip (into your Python; its Scripts/bin must be on PATH):

git clone https://github.com/Jude-A/lucidadl
cd lucidadl
pip install .            # or `pip install -e .` to keep editing the code
playwright install chromium

Open a new terminal afterwards so lucida is picked up. ffmpeg is bundled (imageio-ffmpeg) — nothing to install.

Quick start

lucida setup                                  # once: pass Cloudflare, cache the cookie
lucida                                         # interactive menu (same as `lucida ui`)
lucida track "Red Hot Chili Peppers - Otherside"
lucida album "Red Hot Chili Peppers - Californication" --to mp3 --bitrate 320k -j 8

Prefer a menu? Run lucida with no arguments (or lucida ui): pick an action, type a query/URL, and watch one progress bar per parallel download. Your menu defaults (jobs, service, format, folder) are remembered.

Files land in one fixed folder — ~/Downloads/music by default (not the current directory, so they never scatter). Change it once with lucida config --music "D:/Music" (or the LUCIDADL_MUSIC env var), or per run with -o.

Commands

Singular = ad-hoc (arguments, always downloads). Plural = watchlist (reads a file, skips already-downloaded items — for unattended/scheduled runs).

Command	Input	Dedup
lucida / lucida ui	interactive menu	—
lucida track "<query|url>"	argument(s)	no (force)
lucida album "<query|url>"	argument(s)	no (force)
lucida tracks	./inputs/tracks.txt	yes
lucida albums	./inputs/albums.txt	yes
lucida playlist "<apple music url>"	public playlist	yes
lucida search "<query>"	interactive pick	no
lucida retry	failed list	yes
lucida config	show/set the music folder	—
lucida setup	—	—
lucida doctor	environment check	—

A search takes the best-matching result (title + artist, avoiding remix/cover/karaoke/ live/… unless you ask for them). A playlist/album URL downloads all its tracks.

Options (download commands)

• -j, --jobs N — parallel downloads, 1–20 (default 3).
• -s, --service — qobuz (default) or amazon. If the primary finds nothing, it falls back to the other automatically.
• -F, --format — format requested from lucida (server-side, no bitrate control): original (default) · flac · mp3 · ogg-vorbis · opus · m4a-aac · wav.
• --to — local ffmpeg transcode (recommended for a precise format/bitrate): mp3 · aac/m4a · opus · ogg · flac · wav. Downloads FLAC then converts.
• --bitrate — e.g. 320k, 256k, 192k (for --to).
• --keep-original — keep the source FLAC next to the transcoded file.
• --force — ignore the dedup memory and re-download even items already recorded as done (handy if state.json drifted out of sync).
• --organize / --flat — tag-based Artists/<Artist>/<Album>/ (default) vs everything flat in <music folder>/Music/.
• --country — country code (default US for Qobuz; Amazon needs none).
• -o, --out — output directory for this run (default: the configured music folder, ~/Downloads/music).
• --hidden / --visible — if a Cloudflare refresh is needed, open the window off-screen (--hidden) instead of visible.

Watchlists

Copy the example files and edit them — one item per line (a search artist - title, or a direct URL):

cp inputs/tracks.txt.example inputs/tracks.txt
cp inputs/albums.txt.example inputs/albums.txt

then:

lucida tracks      # downloads everything new, skips what's already done
lucida albums

Playlists

lucida playlist "https://music.apple.com/.../pl.xxxxxxxx" [--dry-run] [-j N]

Give it a playlist link: lucidadl reads the track list (title + artist) straight from the page, then downloads each track through lucida (Qobuz) into Playlists/<playlist name>/. --dry-run just lists them (and writes ./inputs/playlist.txt) without downloading.

Only Apple Music is implemented today — that's what I use. Adding other sources (Spotify, Deezer, Tidal…) is straightforward, and contributions are welcome. The playlist scraping all lives in lucidadl/api.py:

• playlist_tracklist() — picks a scraper based on the link's host.
• applemusic_tracklist() — the working Apple Music scraper (your reference example).
• _scrape_playlist() + _PLAYLIST_SOURCES — a generic scraper with per-site CSS selectors (best-guess starting points for Spotify/Deezer/Tidal), gated behind the _PLAYLIST_OTHERS_ENABLED flag.

To add a source: flip _PLAYLIST_OTHERS_ENABLED = True, fix the selectors for your service in _PLAYLIST_SOURCES, test, and open a PR.

Scheduling / "in the background"

The cookie is cached, so unattended runs open no browser (until it expires). Schedule a watchlist with your OS scheduler. A Windows example is provided in schedule.ps1.

Where files live

• Music: one fixed folder, ~/Downloads/music by default. Set it with lucida config --music "<path>" or the LUCIDADL_MUSIC env var; lucida config (no args) prints every path. Everything is saved here and deduped against here only.
• App data (browser profile, clearance.json, dedup state.json, config.json, run.log, failed.txt): the OS user data dir (%LOCALAPPDATA%\lucidadl on Windows, ~/.local/share/lucidadl on Linux, ~/Library/Application Support/lucidadl on macOS). Override with LUCIDADL_HOME.
• Watchlist inputs (tracks.txt, albums.txt): ./inputs/ next to where you run the command, so you can keep them in your project. Override per command with -f.

Troubleshooting

• Everything lands in Unknown Artist/Unknown Album → mutagen is missing in the Python that runs lucida (tags can't be read). pip install mutagen into that interpreter (it's a declared dependency, so a normal pip install ./pipx install pulls it). lucidadl now prints a warning when it's absent.
• "Cloudflare not cleared" → run lucida setup again (the cached cookie expired).
• "Executable doesn't exist" → run playwright install chromium.
• Search finds nothing → try a direct URL, or -s amazon.
• lucida doctor → checks Python, Playwright, and reachability.

Credits

lucidadl takes inspiration from two existing open-source lucida.to downloaders we looked at while building it:

• lucida-flow — a Python CLI/API that drives lucida.to through browser automation; the starting point for the browser side.
• lucida-downloader — a fast, multithreaded Rust client; the inspiration for downloading many tracks concurrently.

It's a "vibe-coded" project (built quickly, AI-assisted), so expect rough edges — issues and PRs that sharpen or extend it are very welcome.

Contributing

Bug reports and PRs welcome — see CONTRIBUTING.md for the dev setup and how to run the offline self-tests. Notable changes are tracked in CHANGELOG.md.

License

MIT — see LICENSE.