hc-tec/my-collection-skills/media-audio-download

Media Audio Download (Docker)

This skill turns a video URL into an audio file you can feed into an STT skill (e.g. whisper-transcribe-docker).

Supported (best-effort):

• Bilibili video URLs (via yt-dlp)
• XiaoHongShu note URLs (via Playwright -> extract masterUrl -> ffmpeg extract audio)

Quick Start

Build image:

docker build -t moltbot-media-audio-download {baseDir}

Download audio from a URL into a host folder:

mkdir -p out
docker run --rm -v "$PWD/out:/out" moltbot-media-audio-download --url 'https://www.bilibili.com/video/BV...'

Auth via CookieCloud (recommended)

If the page requires login cookies, pass CookieCloud env vars into the container.

CookieCloud server URL (inside Docker):

• Windows/macOS: http://host.docker.internal:8088
• Linux: may require --add-host=host.docker.internal:host-gateway

Example:

docker run --rm -v "$PWD/out:/out" \
  -e COOKIECLOUD_SERVER_URL='http://host.docker.internal:8088' \
  -e COOKIECLOUDUUID='COOKIECLOUD_UUID_HERE' \
  -e COOKIECLOUDPASSWORD='COOKIECLOUD_PASSWORD_HERE' \
  moltbot-media-audio-download --url 'https://www.bilibili.com/video/BV...'

XiaoHongShu notes that require xsec_token

Many XHS notes need an xsec_token to browse. Avoid shell escaping by passing it separately:

docker run --rm -v "$PWD/out:/out" \
  -e COOKIECLOUD_SERVER_URL='http://host.docker.internal:8088' \
  -e COOKIECLOUDUUID='COOKIECLOUD_UUID_HERE' \
  -e COOKIECLOUDPASSWORD='COOKIECLOUD_PASSWORD_HERE' \
  moltbot-media-audio-download --note-id <noteId> --xsec-token <xsec_token>

Output

• Default output dir: /out (mount a host directory there).
• Default format: m4a (good for Whisper/faster-whisper).
• Use --json to get machine-readable output (path, platform, ids).