---

name: iopho-getting-videos description: Download video, audio, subtitles, and metadata from YouTube, Bilibili, Vimeo, and 1800+ platforms allowed-tools: Bash(yt-dlp *), Bash(python3 *), Bash(ffmpeg *), Bash(BBDown *), Bash(lux *), Bash(you-get *) user-invocable: true argument-hint: [--mode video|audio|transcript|metadata|all] [--quality best|720p|480p] [--output DIR] updated: "2026-02-27"

iopho-getting-videos — Cross-Platform Video/Audio/Subtitle Downloader

Download video, audio, subtitles, thumbnails, and metadata from 1800+ platforms with intelligent tool routing and fallback chains.

Prerequisites

Check what's available:

!which yt-dlp 2>/dev/null && echo "yt-dlp: ✓ (primary — covers 1800+ sites)" || echo "yt-dlp: ✗ — install: pip install yt-dlp"

!which ffmpeg 2>/dev/null && echo "ffmpeg: ✓ (needed for merging/conversion)" || echo "ffmpeg: ✗ — install: brew install ffmpeg"

!python3 -c "from youtube_transcript_api import YouTubeTranscriptApi; print('youtube-transcript-api: ✓ (fast YouTube transcripts)')" 2>/dev/null || echo "youtube-transcript-api: ✗ (optional) — install: pip install youtube-transcript-api"

!which BBDown 2>/dev/null && echo "BBDown: ✓ (Bilibili specialist)" || echo "BBDown: ✗ (optional for Bilibili) — install: dotnet tool install -g BBDown"

!which lux 2>/dev/null && echo "lux: ✓ (Chinese platforms)" || echo "lux: ✗ (optional) — install: go install github.com/iawia002/lux@latest"

Minimum: yt-dlp + ffmpeg covers 95% of use cases.

Modes

Mode	What You Get	Primary Tool
video	Video file (best quality)	yt-dlp
audio	Audio-only (MP3/M4A)	yt-dlp -x
transcript	Subtitles/captions text	youtube-transcript-api → yt-dlp --write-subs
metadata	JSON with title, duration, tags, etc.	yt-dlp --dump-json
all	Video + audio + subs + metadata + thumbnail	yt-dlp (combined flags)

Quick Reference

Download video (best quality):

yt-dlp -f "bestvideo[height<=1080]+bestaudio/best" --merge-output-format mp4 -o "$OUTPUT_DIR/%(title)s.%(ext)s" "$URL"

Audio only (MP3):

yt-dlp -x --audio-format mp3 --audio-quality 0 -o "$OUTPUT_DIR/%(title)s.%(ext)s" "$URL"

Transcript only (fastest — YouTube):

python3 -c "
from youtube_transcript_api import YouTubeTranscriptApi
import re, sys
video_id = re.search(r'(?:v=|youtu\.be/)([a-zA-Z0-9_-]{11})', '$URL').group(1)
try:
    transcript = YouTubeTranscriptApi.get_transcript(video_id)
    for entry in transcript:
        m, s = divmod(int(entry['start']), 60)
        print(f'[{m}:{s:02d}] {entry[\"text\"]}')
except Exception:
    transcript = YouTubeTranscriptApi.get_transcript(video_id, languages=['zh-Hans','zh-Hant','ja','ko','en'])
    for entry in transcript:
        m, s = divmod(int(entry['start']), 60)
        print(f'[{m}:{s:02d}] {entry[\"text\"]}')
"

Transcript via yt-dlp (any platform):

yt-dlp --write-subs --write-auto-subs --sub-lang "en,zh-Hans,zh-Hant" --sub-format srt --skip-download -o "$OUTPUT_DIR/%(title)s" "$URL"

Metadata only (no download):

yt-dlp --dump-json "$URL" 2>/dev/null | python3 -c "
import sys, json
v = json.loads(sys.stdin.read())
print(f\"Title: {v['title']}\")
print(f\"Duration: {v.get('duration','?')}s\")
print(f\"Resolution: {v.get('width','?')}x{v.get('height','?')}\")
print(f\"FPS: {v.get('fps','?')}\")
print(f\"Channel: {v.get('channel', v.get('uploader','?'))}\")
print(f\"Upload date: {v.get('upload_date','?')}\")
print(f\"View count: {v.get('view_count','?')}\")
print(f\"URL: {v.get('webpage_url', v.get('url','?'))}\")
"

Everything (video + subs + metadata + thumbnail):

yt-dlp -f "bestvideo[height<=1080]+bestaudio/best" --merge-output-format mp4 \
  --write-subs --write-auto-subs --sub-lang "en,zh-Hans,zh-Hant" --sub-format srt \
  --write-info-json --write-thumbnail \
  -o "$OUTPUT_DIR/%(title)s.%(ext)s" "$URL"

Execute

Parse the user's arguments and run the appropriate download:

# Parse arguments from: $ARGUMENTS
# Expected: <url> [--mode video|audio|transcript|metadata|all] [--quality best|720p|480p] [--output DIR]
#
# Default: --mode video --quality best --output ./
#
# The agent should:
# 1. Detect platform from URL domain
# 2. Choose optimal tool (see Routing Logic below)
# 3. Run download with appropriate flags
# 4. Report: file path, size, duration, format

Routing Logic

The agent should auto-detect the best tool based on URL:

URL contains youtube.com / youtu.be:
  transcript-only? → youtube-transcript-api (fastest, no download)
  else → yt-dlp

URL contains bilibili.com / b23.tv:
  yt-dlp fails? → BBDown (handles multi-part, interactive videos)
  transcript-only? → yt-dlp --write-subs (Bilibili CC subs)

URL contains youku/iqiyi/qq.com/mgtv:
  yt-dlp fails? → lux (Chinese platform specialist)

URL contains vimeo.com:
  yt-dlp (works for most public videos)
  private/embedded? → try yt-dlp --referer <page_url>

URL contains tiktok.com / douyin.com:
  yt-dlp (usually works)

URL contains twitter.com / x.com:
  yt-dlp (usually works)

Any other URL:
  yt-dlp (covers 1800+ extractors)
  yt-dlp fails? → you-get (80+ sites, different extraction logic)

Fallback Chain

When the primary tool fails, try the next option:

1. yt-dlp (standard)
   ↓ fails?
2. yt-dlp --cookies-from-browser chrome (auth/age-restricted)
   ↓ fails?
3. yt-dlp --geo-bypass (geo-restricted)
   ↓ fails?
4. Platform-specific tool (BBDown / lux / you-get)
   ↓ fails?
5. Report error with diagnostic info

Advanced Usage

List available formats (before downloading):

yt-dlp --list-formats "$URL"

Download specific format (by format ID):

yt-dlp -f "FORMAT_ID" -o "$OUTPUT_DIR/%(title)s.%(ext)s" "$URL"

Download with aria2c acceleration (2-5x faster):

yt-dlp --downloader aria2c --downloader-args aria2c:"-x 16 -s 16" "$URL"

Bilibili with BBDown (when yt-dlp fails):

BBDown "$URL" -tv --work-dir "$OUTPUT_DIR"

Chinese platforms with lux:

lux -o "$OUTPUT_DIR" "$URL"

Extract audio from already-downloaded video:

ffmpeg -i "$VIDEO_FILE" -vn -acodec libmp3lame -q:a 2 "${VIDEO_FILE%.*}.mp3"

Local transcription (when no subtitles exist):

# Requires: pip install faster-whisper
python3 -c "
from faster_whisper import WhisperModel
model = WhisperModel('base', compute_type='int8')
segments, _ = model.transcribe('$AUDIO_FILE')
for seg in segments:
    m, s = divmod(int(seg.start), 60)
    print(f'[{m}:{s:02d}] {seg.text}')
"

Combine with Other iopho Skills

# 1. Search for videos
/iopho-searching-videos SaaS product demo under 30s

# 2. Download the best match
/iopho-getting-videos https://youtube.com/watch?v=... --mode all

# 3. Analyze into storyboard
/iopho-analyzing-videos ./video.mp4

Tips

• For YouTube shorts, the same URL format works with yt-dlp
• Use --match-filter "duration<60" during search (in iopho-searching-videos) to pre-filter
• --write-info-json creates a .info.json file that iopho-analyzing-videos can use for metadata
• For Bilibili, login cookies help access higher quality: yt-dlp --cookies-from-browser chrome
• Use --no-playlist to download a single video from a playlist URL