New tool that classifies URLs by domain and delegates to pos media mp3 (audio) or pos media mp4 (video). Listener gains URL detection step between prefix map and AI bridge — bare URLs auto-download. Domain rules: music.youtube.com/soundcloud/bandcamp → mp3, youtube/vimeo/twitch → mp4 --best, unknown → configurable default. 28 test cases / 70 assertions / 0 failed. make gen && make check green, make lint 0 FAIL / 0 WARN.
12 KiB
How-To: pos media
Download audio and video from the web via yt-dlp, auto-classify URLs,
sync your library to a USB stick, and keep YouTube channels incrementally
up to date.
Tools: grab, mp3, mp4, sync, ytsync.
| Tool | What it does |
|---|---|
pos media grab |
Auto-classify URL and download as audio or video |
pos media mp3 |
Download audio, convert to MP3 |
pos media mp4 |
Download video with smart/interactive format selection |
pos media sync |
Incrementally copy ~/Music onto a USB stick (mp3/mp4) |
pos media ytsync |
Track YouTube channels/playlists and download only new videos into ~/Videos |
Requires yt-dlp and ffmpeg (sudo apt install yt-dlp ffmpeg); the tools
fail with a clean error message instead of a raw command not found if either
is missing.
pos media mp3 — audio as MP3
pos media mp3 <url>
Extracts and converts the audio track to MP3 in ~/Music/. With --by-artist
files land in ~/Music/<artist>/<title>.mp3 (falls back to the uploader name
when there's no artist tag), so a library stays organized.
pos media mp3 https://youtube.com/watch?v=dQw4w9WgXcQ
pos media mp3 --by-artist https://youtu.be/dQw4w9WgXcQ
MP3s are encoded at best quality with title/artist/album/date/chapters embedded
(--embed-metadata --embed-chapters, artist falls back to uploader) plus the
cover art as a JPEG thumbnail. Existing files are never overwritten.
| Flag | Meaning |
|---|---|
-o, --output <dir> |
Output directory (default ~/Music) |
--by-artist |
Organize as <dir>/<artist>/<title>.mp3 |
--no-playlist |
Download only the single video, not the playlist |
--cookies <file> |
Netscape cookies.txt for age-gated content |
--dry-run |
Print the yt-dlp command without running it |
Recipe: batch — loop over a list of URLs:
while read -r url; do pos media mp3 --by-artist "$url"; done < urls.txt
pos media mp4 — video with smart format choice
pos media mp4 <url>
Without a format flag, the available formats are shown as a short curated
list (grouped [audio] / [video] / [combo], with the raw dump's clutter
dropped) and you pick one — the id is validated before downloading. Entering
nothing (or best) picks the best video+audio automatically.
Non-interactive (scripting-friendly):
| Flag | Meaning |
|---|---|
-f, --format <id> |
Download that format id directly (no prompt) |
--best |
Best video + audio, no prompt |
--worst |
Lowest quality, no prompt |
-o, --output <dir> |
Output directory (default ~/Videos) |
--no-playlist |
Download only the single video |
--cookies <file> |
Netscape cookies.txt for age-gated content |
--dry-run |
Print the yt-dlp command without running it |
pos media mp4 --best https://youtube.com/watch?v=dQw4w9WgXcQ
pos media mp4 -f 22 https://youtube.com/watch?v=dQw4w9WgXcQ
Videos merge to MP4 with metadata, chapters, subtitles (all languages) and the
thumbnail embedded (--embed-metadata --embed-chapters --embed-subs --sub-langs all --embed-thumbnail); existing files are never overwritten.
Recipe: grab a 4K stream for later — --best already picks the best
video+audio and merges them.
pos media grab — auto-classify URL and download
pos media grab <url>
Smart URL classifier that routes to pos media mp3 or pos media mp4
automatically based on the domain. Send a URL from your phone via Telegram and
the bot downloads it to the right place without you thinking about it.
Classification rules:
| Domain | Routes to | Why |
|---|---|---|
music.youtube.com |
mp3 | Audio streaming |
soundcloud.com |
mp3 | Audio-first platform |
bandcamp.com |
mp3 | Audio-first platform |
youtube.com / youtu.be |
mp4 | Video content |
vimeo.com / twitch.tv |
mp4 | Video platforms |
| Everything else | mp4 (default) | Safe fallback |
Override the classification with --audio or --video. The default for
unrecognized domains is video — change it with pos config grab or set
GRAB_DEFAULT=audio in ~/.config/linux_post_install/grab.env.
pos media grab https://music.youtube.com/watch?v=abc # → ~/Music
pos media grab https://youtube.com/watch?v=xyz # → ~/Videos
pos media grab --audio https://vimeo.com/123 # force mp3
pos media grab --worst https://youtu.be/abc # lowest quality
pos media grab --dry-run https://soundcloud.com/artist/track # preview only
Non-interactive by design — pos media mp4 receives --best by default so it
never prompts for a format (critical for Telegram bot context where there's no
TTY). Pass --worst if you want the smallest file.
| Flag | Meaning |
|---|---|
--audio |
Force audio (mp3) download |
--video |
Force video (mp4) download |
--best |
Best quality for video (default) |
--worst |
Lowest quality for video |
-o, --output <dir> |
Output directory (passed to mp3/mp4) |
--no-playlist |
Download only the single video |
--cookies <file> |
Netscape cookies.txt for age-gated content |
--dry-run |
Print the command that would run, don't execute |
pos media sync — music onto a USB stick
pos media sync # copy everything (mp3 + mp4) from ~/Music to the stick
pos media sync --mp3 # only the .mp3 files
pos media sync --mp4 # only the .mp4 files
Detects connected USB storage exactly like pos system backup (same shared
lib/usb-lib.sh): a plugged-in but unmounted stick is offered a mount first
(/media/<label>, world-writable, mirrors usb-automount), multiple sticks
are listed for you to pick, and if nothing is plugged in it re-scans after you
press Enter. Files are mirrored into <usb>/Music/ (change with
MEDIA_SYNC_DEST) preserving the artist/album tree.
Sync semantics — add/update only, never delete. Files missing on the stick
are copied; files whose size or mtime changed are overwritten; everything
identical is skipped. Files on the stick that are no longer in the source are
left alone — a playback stick can never lose files to a mirror mistake.
Copies keep the source timestamps (cp --preserve=timestamps), so a re-run is
a no-op. Preview before copying with --dry-run:
pos media sync --mp4 --dry-run # shows "would copy" list + counts, copies nothing
The source folder may be a symlink to a library elsewhere
(~/Music -> /mnt/data/music) — it is followed, the artist/album tree is
mirrored under the symlink's target.
Target picking. Mounted USB partitions are listed with size, label and
filesystem (single candidate → confirm prompt; several → numbered picker, one
row per partition). EFI system partitions (e.g. a Ventoy stick's VTOYEFI,
32 MB) are never offered as a target — they are boot machinery, not
storage. Before any copy the tool verifies the payload fits (df vs the
exact bytes to copy) and fails fast with Not enough free space on … instead
of dying mid-copy with No space left on device.
| Flag | Meaning |
|---|---|
--mp3 |
Sync only *.mp3 (neither flag = both) |
--mp4 |
Sync only *.mp4 (neither flag = both) |
--source <dir> |
Source folder (default ~/Music) |
--dry-run |
Preview what would be copied, copy nothing |
When it finishes it announces the result via lib/notify.sh
(Music sync completed: N added, M updated → <usb>/Music).
Recipes:
pos media sync # keep the car stick up to date (both formats)
pos media sync --mp3 --dry-run # check what a new batch will bring first
pos media sync --source /data/Music # sync a library that lives elsewhere
Config (all in ~/.config/linux_post_install/system.env or exported):
MEDIA_SYNC_SOURCE (default $HOME/Music), MEDIA_SYNC_DEST (default
Music), plus the shared USB_MOUNT_BASE / USB_BYID seams.
pos media ytsync — incremental YouTube channel sync
pos media ytsync # interactive
pos media ytsync add https://youtube.com/@SomeChannel
pos media ytsync sync # fetch new videos from every source
pos media ytsync list
pos media ytsync remove "Some Channel"
The first run asks for a channel or playlist URL, shows what was resolved
(Resolved : Linus Tech Tips (channel · 2140 videos) + the target library path),
and downloads everything. Every later run probes the same URL, diffs it against a
per-source download archive, and fetches only new videos — one yt-dlp call per
video (bestvideo*+bestaudio/best merged to MP4 with metadata, chapters and the
thumbnail embedded; existing files are never overwritten).
Where files land:
~/Videos/<channel>/<title>.mp4 # channels & single videos
~/Videos/<channel>/<playlist>/<NNN> - <title>.mp4 # playlist sources (playlist order)
| Command | Meaning |
|---|---|
add [url] |
Register a source + first download (asks for the URL if omitted) |
sync [name] |
Incremental pass — all tracked sources, or one by slug/display name |
list |
Tracked sources table: type, archived count, last sync date, destination |
remove <name> |
Stop tracking. Downloaded files AND the archive are kept |
Notes:
- A watch link with both
?v=and&list=downloads only that single video, never the whole playlist;youtu.be/<id>short links count as watch links too. - Retitled/renamed videos keep their local filenames (the archive is keyed by video id); inserting a video mid-playlist shifts future numbering only.
- Members-only / age-gated videos are skipped with a count ("N videos require
sign-in"). Escape hatch: put
YTSYNC_EXTRA_ARGS=--cookies-from-browser firefox(or--cookies <file>) into the config file below. - State lives outside
~/Videos, in~/.local/share/linux_post_install/ytsync/(registry, per-source archives, run history) — the media tree stays pure media.
Recipe: daily automation — no built-in timers; use the shared scheduler:
pos system schedule config # name: ytsync
# INTERVAL=daily (or hourly / weekly / OnCalendar=…)
# NOTIFY=never # ytsync sends its own digest; don't double-notify
# MSG="ytSync"
# COMMAND=pos media ytsync sync
pos system schedule enable ytsync
pos system schedule run ytsync # test once, right now
ytsync never prompts on sync, so the job is timer-safe by construction. It
notifies via lib/notify.sh only when something happened (new videos or failures);
a scheduled no-op run stays silent — hence NOTIFY=never on the job to avoid
double alerts.
Previewing: --dry-run works on add and sync — real probe + plan block
(New : 12 would be downloaded (800 already present) + example filenames), zero
writes anywhere:
pos media ytsync sync --dry-run
pos media ytsync add https://youtube.com/@SomeChannel --dry-run
Config (~/.config/linux_post_install/ytsync.env, chmod 600 — materialize/edit via
pos config ytsync; exported environment wins):
| Key | Default | Meaning |
|---|---|---|
YTSYNC_VIDEOS_DIR |
$HOME/Videos |
Videos root for synced sources |
YTSYNC_EXTRA_ARGS |
(empty) | Extra yt-dlp flags appended to every download call (cookies, format overrides, …) |
Troubleshooting
- Format list is empty / download fails → the site or age-gate requires
cookies; pass
--cookies ~/cookies.txt(export it from your browser), or update yt-dlp (sudo apt upgrade yt-dlp). - Error about a missing postprocessor →
sudo apt install ffmpeg. --by-artistleaves files loose → the source has no artist/uploader tag; it falls back to the uploader name in the artist slot.- Very large downloads: ensure free space; files land in
~/Music/~/Videos(or your-odirectory). - ytsync: "could not reach YouTube" at add → connectivity/DNS; the interactive flow re-prompts up to 3×. "source not found or private" → wrong handle or a deleted/private source.
- ytsync: many "exists, kept" warnings after a YouTube-side change → the
channel's internal id changed, so re-adding computed a new archive. Rename the
old archive file (
~/.local/share/linux_post_install/ytsync/archive/<old>.txt) to the new slug before syncing (seetools-docs/ytsync.md). - ytsync: "disk full — stopping … mid-run" → free space on the videos volume and
run
pos media ytsync syncagain; unfetched videos simply stay "new".
Related
- Reference: DOC/POS.md → media