xiaoyuzhou-to-audio

xiaoyuzhou-to-audio

Download up to 6 小宇宙 (Xiaoyuzhou FM) episodes to local .m4a files in parallel, using aria2c with up to 6 connections per file. Self-contained.

Transcription is handled separately. After download, point the user at the media-to-srt skill if they want a transcript.

Limits

• Max 6 URLs per invocation. If the user provides more, take the first 6 and tell them to re-run for the rest. Beyond 6 the CDN starts throttling and progress reporting becomes unreadable.

Prerequisites — install if missing

Check both tools first:

command -v aria2c ffprobe curl

If aria2c is missing, install it before proceeding — do not silently fall back to curl:

brew install aria2

If ffprobe is missing:

brew install ffmpeg

After install, re-run the command -v check to confirm. If brew itself is unavailable, stop and tell the user to install Homebrew (https://brew.sh).

Workflow

Work in ~/Downloads/ by default.

1. Resolve audio URL and title for each episode

For each URL the user provided (loop, max 6):

HTML=$(curl -s -A "Mozilla/5.0" "<EPISODE_URL>")
AUDIO_URL=$(echo "$HTML" | grep -oE 'https://media\.xyzcdn\.net/[^"]*\.m4a' | head -1)
RAW_TITLE=$(echo "$HTML" | grep -oE '<title>[^<]+</title>' | head -1 | sed -E 's/<\/?title>//g')

If AUDIO_URL is empty for any episode, report that one as failed and continue with the rest.

Sanitize RAW_TITLE into $PREFIX: strip the trailing - <播客名> | 小宇宙 - 听播客，上小宇宙, replace / and whitespace runs with _. Keep it short and shell-safe. Make sure the 6 prefixes are unique — if two collide, suffix with _2, _3, etc.

2. Download all episodes in parallel via aria2c

Build an aria2c input file (one URL + filename per pair), then run a single aria2c that downloads everything concurrently:

cd ~/Downloads
cat > .xyz_input.txt <<EOF
<AUDIO_URL_1>
  out=<PREFIX_1>.m4a
<AUDIO_URL_2>
  out=<PREFIX_2>.m4a
...
EOF

aria2c -i .xyz_input.txt \
  -j 6 -x 6 -s 6 -k 1M \
  --user-agent="Mozilla/5.0" \
  --console-log-level=warn --summary-interval=5 \
  --auto-file-renaming=false --allow-overwrite=false

Flags:

• -j 6 up to 6 files in parallel
• -x 6 -s 6 6 connections per file, 6 segments (CDN-friendly cap)
• -k 1M 1MB segment size
• --summary-interval=5 prints a progress summary every 5s (use this as the live progress, no separate Monitor needed for short jobs)
• --allow-overwrite=false skip files that already exist (idempotent re-runs)

For long downloads (>5 min expected), launch aria2c with run_in_background: true and use a Monitor that tails the log every 15–30s.

aria2c handles resume automatically — if a download is interrupted, just re-run the same command.

3. Verify each file

For each $PREFIX.m4a produced:

sz=$(stat -f%z ~/Downloads/$PREFIX.m4a)
dur=$(ffprobe -v error -show_entries format=duration -of default=noprint_wrappers=1:nokey=1 ~/Downloads/$PREFIX.m4a)
echo "$PREFIX: $((sz/1024/1024)) MB, ${dur}s"

aria2c exit code 0 means all files downloaded completely; you only need ffprobe to surface the duration to the user. If exit code ≠ 0, list which files failed (aria2c prints a summary at the end).

4. Report

Per-episode line: <title> — <MB> MB, <h:mm:ss>, <output path>. Then list any failures.

Delete .xyz_input.txt when done.

Notes & gotchas

• The CDN host is media.xyzcdn.net. URLs are public; no cookies/auth needed.
• Episode title sanitization: typical pattern is <EP title> - <show name> | 小宇宙 - 听播客，上小宇宙 — strip everything from - onward (or just the | 小宇宙... tail) to keep $PREFIX short.
• aria2c is dramatically faster than curl on xyzcdn (typical 30–50 MB/s vs 3–8 MB/s single-connection).
• If the user provides >6 URLs, take the first 6, finish them, then tell the user to re-invoke for the remaining ones.