media-fetch

name	media-fetch
user-invocable	false
description	Download video/audio from 1000+ sites (YouTube, Twitter/X, Spotify, TikTok, etc.), extract audio, download subtitles, transcribe speech to text, and search for videos. Use when asked to download, summarize, or transcribe any video/audio URL, or when asked to search YouTube.
triggers	["download video","download audio","youtube","transcribe video","summarize this video","get subtitles","yt-dlp","media download","podcast","search youtube","watch this"]
category	media
maturity	stable
tags	["yt-dlp","ffmpeg","whisper","transcription","youtube"]

Media Fetch

Download, extract, and transcribe media from 1000+ sites. Built on yt-dlp + ffmpeg + Whisper.

Scripts: ${CLAUDE_SKILLS_DIR:-$HOME/.claude-agent/.claude/skills}/media-fetch/scripts/ Downloads: ~/.media-fetch/downloads/ Transcripts: ~/.media-fetch/transcripts/ Metadata: ~/.media-fetch/metadata/

Quick Reference

S=${CLAUDE_SKILLS_DIR:-$HOME/.claude-agent/.claude/skills}/media-fetch/scripts

# Download video (best quality)
python3 $S/fetch.py URL

# Audio only (MP3)
python3 $S/fetch.py URL --audio

# Download + transcribe (the "summarize this video" workflow)
python3 $S/fetch.py URL --transcript

# Subtitles only (fastest — no download needed)
python3 $S/fetch.py URL --subs

# Video info (no download)
python3 $S/fetch.py URL --info

# Search YouTube
python3 $S/fetch.py --search "prop trading strategies"

# Extract audio from local file
python3 $S/batch.py --file video.mp4 --extract-audio

# Transcribe local audio/video
python3 $S/batch.py --file audio.mp3 --transcribe

---

Commands — fetch.py

Video Info (No Download)

python3 fetch.py URL --info
python3 fetch.py URL --info --json    # Machine-readable

Shows: title, channel, duration, views, likes, description, chapters, available formats, subtitle languages, best video/audio quality.

Download Video

python3 fetch.py URL                       # Best quality MP4
python3 fetch.py URL --quality 720         # Max 720p
python3 fetch.py URL --quality 480         # Max 480p
python3 fetch.py URL -o /tmp/videos        # Custom output dir
python3 fetch.py URL --format "bestvideo[height<=1080]+bestaudio"  # Custom format

Download Audio Only

python3 fetch.py URL --audio               # MP3, best quality
python3 fetch.py URL --audio --clip 5:00-10:00   # Extract clip

Download Subtitles Only

The fastest way to get video content — no media download needed:

python3 fetch.py URL --subs                    # English subtitles
python3 fetch.py URL --subs --sub-lang en,no   # English + Norwegian
python3 fetch.py URL --subs --sub-lang auto    # Auto-generated captions

Returns cleaned text (timestamps/formatting stripped). Works with:

• YouTube (manual + auto-generated captions)
• Most major platforms with subtitle support

Download + Transcribe

The full "summarize this video" pipeline:

python3 fetch.py URL --transcript              # Auto: tries subs → Whisper API → local
python3 fetch.py URL --transcript --transcribe-method whisper-api
python3 fetch.py URL --transcript --transcribe-method subtitles
python3 fetch.py URL --transcript --language no    # Norwegian audio

Transcription methods (in --transcribe-method):

Method	Speed	Cost	Quality	Requirements
subtitles	⚡ Instant	Free	Good (if available)	Video must have subs/captions
whisper-api	🏃 Fast	~$0.006/min	Excellent	OPENAI_API_KEY env var
whisper-local	🐌 Slow	Free	Good	pip install openai-whisper + GPU recommended
auto	Varies	Varies	Best available	Tries subs → API → local

Search

python3 fetch.py --search "forex trading NFP strategy"
python3 fetch.py --search "react server components tutorial"
python3 fetch.py --search "home espresso machine review" --json

Clip Extraction

Download only a portion of a video:

python3 fetch.py URL --clip 1:30-5:00          # Minutes:seconds
python3 fetch.py URL --clip 0:00-0:30 --audio  # First 30s audio

Playlist Download

python3 fetch.py URL --playlist                    # Full playlist
python3 fetch.py URL --playlist --max-playlist 5   # First 5 items
python3 fetch.py URL --playlist --audio            # Audio from all

---

Commands — batch.py

Process URL List

Create a text file with one URL per line:

# urls.txt
https://youtube.com/watch?v=abc
https://youtube.com/watch?v=def
# Comments and blank lines are ignored
https://vimeo.com/123456

python3 batch.py urls.txt                        # Download all
python3 batch.py urls.txt --audio                # Audio only
python3 batch.py urls.txt --transcript           # Download + transcribe all
python3 batch.py urls.txt -o /tmp/batch-output   # Custom output dir

Process Local Files

python3 batch.py --file video.mp4 --extract-audio           # → video.mp3
python3 batch.py --file audio.mp3 --transcribe               # → transcript text
python3 batch.py --file audio.mp3 --transcribe --language no # Norwegian
python3 batch.py --file video.mp4 --file-info                # ffprobe metadata

---

Supported Sites

yt-dlp supports 1000+ sites including:

Category	Sites
Video	YouTube, Vimeo, Dailymotion, Twitch, Rumble, Odysee
Social	Twitter/X, TikTok, Instagram, Reddit, Facebook
Music	SoundCloud, Bandcamp, Mixcloud
Podcasts	Spotify (some), Apple Podcasts (some), RSS feeds
News	BBC, CNN, Reuters, Bloomberg
Education	Coursera, Udemy, Khan Academy
Live	Twitch live, YouTube live
Other	PeerTube, BitChute, Archive.org, and 1000+ more

Full list: yt-dlp --list-extractors

---

Agent Workflows

"Summarize this YouTube video"

# Step 1: Get subtitles (fastest path)
python3 fetch.py URL --subs

# Step 2: If subs available, summarize the text directly
# If not, download audio and transcribe:
python3 fetch.py URL --transcript

# Step 3: Read the transcript and summarize
cat ~/.media-fetch/transcripts/TITLE.txt

"Find and summarize a video about X"

# Step 1: Search
python3 fetch.py --search "topic query"

# Step 2: Pick best result, get transcript
python3 fetch.py CHOSEN_URL --transcript

"Download this podcast episode"

python3 fetch.py PODCAST_URL --audio

"Get the audio from this meeting recording"

python3 batch.py --file recording.mp4 --extract-audio
python3 batch.py --file recording.mp3 --transcribe

"Download all videos from this playlist"

python3 fetch.py PLAYLIST_URL --playlist --audio --max-playlist 20

---

Authentication / Cookies

Some sites require login. Export browser cookies:

# Option 1: Use browser cookies directly
python3 fetch.py URL --extra --cookies-from-browser chrome

# Option 2: Export cookies file (Netscape format)
python3 fetch.py URL --cookies ~/cookies.txt

---

Dependencies

Installed:

• yt-dlp (2026.03.03) — media downloader
• ffmpeg (6.1.1) — audio/video processing
• openai SDK — Whisper API transcription

Optional (not installed):

• openai-whisper — local transcription: pip3 install --break-system-packages openai-whisper (Requires ~1GB disk + CUDA GPU for reasonable speed)

---

Storage

Directory	Contents	Cleanup
~/.media-fetch/downloads/	Downloaded video/audio files	Delete manually
~/.media-fetch/transcripts/	Text transcriptions + SRT files	Keep for reference
~/.media-fetch/metadata/	JSON metadata from --info	Auto-cleaned

To clean old downloads:

find ~/.media-fetch/downloads -mtime +7 -delete    # Older than 7 days