agentic/code/frameworks/media-curator/skills/cover-art-embedding/SKILL.md
Patterns for finding, processing, and embedding cover artwork into media files
npx skillsauth add jmagly/aiwg Cover Art EmbeddingInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
Concrete patterns for finding high-quality cover artwork, processing it for optimal size and format, and embedding it into audio and video files.
MusicBrainz Cover Art Archive (CAA) is the authoritative source for album artwork.
Get Front Cover by Release MBID:
# Download front cover for a release
curl -s "https://coverartarchive.org/release/{mbid}/front" -o cover.jpg
# Example with real MBID
curl -s "https://coverartarchive.org/release/5c004fe3-6e96-3f43-9a85-16e5be8805fa/front" -o blue-cover.jpg
Check if Cover Exists:
# HEAD request to check without downloading
if curl -s -I "https://coverartarchive.org/release/${mbid}/front" | grep -q "200 OK"; then
echo "Cover available"
else
echo "No cover found"
fi
Get All Available Images:
# Get metadata for all images
curl -s "https://coverartarchive.org/release/${mbid}" | jq .
# Extract URLs for all front covers
curl -s "https://coverartarchive.org/release/${mbid}" | \
jq -r '.images[] | select(.front == true) | .image'
# Download highest resolution front cover
url=$(curl -s "https://coverartarchive.org/release/${mbid}" | \
jq -r '.images[] | select(.front == true) | .image' | head -1)
curl -s "$url" -o cover.jpg
fanart.tv provides high-quality artist images, album covers, and logos.
Get Artist Images:
# Requires API key from https://fanart.tv/get-an-api-key/
FANART_API_KEY="your-api-key"
# Get all artist artwork
curl -s "https://webservice.fanart.tv/v3/music/${mbid}?api_key=${FANART_API_KEY}"
# Extract album cover URLs
curl -s "https://webservice.fanart.tv/v3/music/${mbid}?api_key=${FANART_API_KEY}" | \
jq -r '.albums."'"$album_mbid"'".albumcover[0].url'
# Extract artist background
curl -s "https://webservice.fanart.tv/v3/music/${mbid}?api_key=${FANART_API_KEY}" | \
jq -r '.artistbackground[0].url'
# Extract artist logo
curl -s "https://webservice.fanart.tv/v3/music/${mbid}?api_key=${FANART_API_KEY}" | \
jq -r '.hdmusiclogo[0].url'
Download Artwork:
# Download album cover
url=$(curl -s "https://webservice.fanart.tv/v3/music/${mbid}?api_key=${FANART_API_KEY}" | \
jq -r '.albums."'"$album_mbid"'".albumcover[0].url')
curl -s "$url" -o cover.jpg
Extract cover artwork from video frames.
Using yt-dlp (if video was downloaded with thumbnail):
# Download video with thumbnail
yt-dlp --write-thumbnail --convert-thumbnails jpg "https://youtube.com/watch?v=..."
# Result: video.jpg alongside video.mp4
Using ffmpeg (extract from video):
# Extract single frame at specific time
ffmpeg -i input.mp4 -ss 00:00:10 -frames:v 1 -q:v 2 thumbnail.jpg
# Extract I-frame (best quality)
ffmpeg -i input.mp4 -vf "select=eq(pict_type\,I)" -frames:v 1 -q:v 2 thumbnail.jpg
# Extract frame at 25% into video
duration=$(ffprobe -v error -show_entries format=duration -of default=noprint_wrappers=1:nokey=1 input.mp4)
timestamp=$(echo "$duration * 0.25" | bc)
ffmpeg -i input.mp4 -ss "$timestamp" -frames:v 1 -q:v 2 thumbnail.jpg
# Extract highest quality frame from first minute
ffmpeg -i input.mp4 -t 60 -vf "select=eq(pict_type\,I)" -vsync vfr -q:v 1 frames-%03d.jpg
# Then manually select best frame
Scan for existing artwork in standard locations:
# Check for common filenames in album directory
for name in cover.jpg Cover.jpg folder.jpg Folder.jpg albumart.jpg AlbumArt.jpg; do
if [ -f "$name" ]; then
echo "Found: $name"
break
fi
done
# Search parent directory
if [ -f "../cover.jpg" ]; then
cp "../cover.jpg" .
fi
# Find largest image file (likely the cover)
largest=$(find . -maxdepth 1 -type f \( -iname "*.jpg" -o -iname "*.png" \) -exec ls -s {} \; | \
sort -rn | head -1 | awk '{print $2}')
echo "Largest image: $largest"
Reduce image size for faster loading and smaller file sizes:
# Install ImageMagick
brew install imagemagick # macOS
sudo apt install imagemagick # Debian/Ubuntu
# Resize to 1200x1200 max (preserves aspect ratio)
convert input.jpg -resize 1200x1200 output.jpg
# Resize and optimize quality (85% JPEG quality)
convert input.jpg -resize 1200x1200 -quality 85 output.jpg
# Create thumbnail (300x300)
convert input.jpg -resize 300x300 thumbnail.jpg
# Batch resize all images in directory
for img in *.jpg; do
convert "$img" -resize 1200x1200 -quality 85 "${img%.jpg}-resized.jpg"
done
Convert between image formats:
# PNG to JPEG
convert input.png output.jpg
# WebP to JPEG
convert input.webp output.jpg
# JPEG to PNG (lossless)
convert input.jpg output.png
# Batch convert all PNG to JPEG
for img in *.png; do
convert "$img" "${img%.png}.jpg"
done
Crop non-square images to square aspect ratio:
# Center crop to square
convert input.jpg -gravity center -crop 1:1 +repage output.jpg
# Crop from top-left
convert input.jpg -crop 1200x1200+0+0 +repage output.jpg
# Smart crop (removes boring edges)
convert input.jpg -define trim:percent-background=0% -trim +repage -resize 1200x1200 output.jpg
Strip EXIF data before embedding:
# Remove all metadata
convert input.jpg -strip output.jpg
# Or use exiftool
exiftool -all= input.jpg
# Batch strip metadata
for img in *.jpg; do
convert "$img" -strip "${img%.jpg}-clean.jpg"
done
# Embed cover art from JPEG file
opustags input.opus --set-cover cover.jpg -o output.opus
# In-place embedding
opustags -i input.opus --set-cover cover.jpg
# Verify embedding
opustags input.opus | grep -i picture
# Embed cover as attached picture
ffmpeg -i input.mp4 -i cover.jpg \
-map 0 -map 1 \
-c copy \
-disposition:v:1 attached_pic \
output.mp4
# Embed with metadata
ffmpeg -i input.mp4 -i cover.jpg \
-map 0 -map 1 \
-c copy \
-disposition:v:1 attached_pic \
-metadata:s:v:1 title="Album cover" \
-metadata:s:v:1 comment="Cover (front)" \
output.mp4
# Verify embedding
ffprobe -v error -select_streams v -show_entries stream=codec_name,codec_type,disposition input.mp4
# Embed cover with ID3v2.3
ffmpeg -i input.mp3 -i cover.jpg \
-map 0 -map 1 \
-c copy \
-id3v2_version 3 \
-metadata:s:v title="Album cover" \
-metadata:s:v comment="Cover (front)" \
output.mp3
# Verify embedding
ffmpeg -i input.mp3 -an -vcodec copy test-extracted-cover.jpg
# Embed cover into FLAC
ffmpeg -i input.flac -i cover.jpg \
-map 0 -map 1 \
-c copy \
-disposition:v:0 attached_pic \
output.flac
# Using metaflac (alternative)
metaflac --import-picture-from=cover.jpg input.flac
# Opus files in directory
for file in *.opus; do
opustags -i "$file" --set-cover cover.jpg
echo "Embedded: $file"
done
# MP3 files in directory
for file in *.mp3; do
ffmpeg -i "$file" -i cover.jpg \
-map 0 -map 1 -c copy \
-id3v2_version 3 \
-metadata:s:v title="Album cover" \
"${file%.mp3}-covered.mp3"
mv "${file%.mp3}-covered.mp3" "$file"
done
# Process album directories
for album_dir in */; do
# Find cover in album directory
cover=$(find "$album_dir" -maxdepth 1 -iname "cover.jpg" | head -1)
if [ -n "$cover" ]; then
# Embed into all Opus files in this album
for file in "$album_dir"/*.opus; do
opustags -i "$file" --set-cover "$cover"
done
echo "Processed: $album_dir"
else
echo "No cover found: $album_dir"
fi
done
#!/bin/bash
# Process all Opus files, download covers, embed
for file in *.opus; do
# Extract artist and album from tags
artist=$(opustags "$file" | grep "ARTIST=" | cut -d= -f2)
album=$(opustags "$file" | grep "ALBUM=" | cut -d= -f2)
# Query MusicBrainz for release MBID
artist_enc=$(echo "$artist" | jq -sRr @uri)
album_enc=$(echo "$album" | jq -sRr @uri)
mbdata=$(curl -s "https://musicbrainz.org/ws/2/release/?query=artist:${artist_enc}%20AND%20release:${album_enc}&fmt=json" \
-H "User-Agent: MediaCurator/1.0")
mbid=$(echo "$mbdata" | jq -r '.releases[0].id')
# Download cover
if curl -s "https://coverartarchive.org/release/${mbid}/front" -o "/tmp/cover-${mbid}.jpg"; then
# Embed cover
opustags -i "$file" --set-cover "/tmp/cover-${mbid}.jpg"
echo "Embedded cover: $file"
else
echo "No cover found: $file"
fi
# Rate limit
sleep 1
done
From field testing: use a single artwork/ directory for all artwork files, not scattered copies.
artwork/
├── albums/
│ ├── blue.jpg
│ ├── court-and-spark.jpg
│ ├── hejira.jpg
│ └── clouds.jpg
├── artists/
│ ├── joni-mitchell-1970.jpg
│ ├── joni-mitchell-1976.jpg
│ └── joni-mitchell-2000.jpg
├── live/
│ ├── isle-of-wight-1970.jpg
│ └── carnegie-hall-1969.jpg
├── logos/
│ └── joni-mitchell-logo.png
├── promotional/
│ ├── blue-era-promo.jpg
│ └── court-and-spark-promo.jpg
└── ATTRIBUTION.md
# Create structure
mkdir -p artwork/{albums,artists,live,logos,promotional}
# Download album covers
cd artwork/albums
curl -s "https://coverartarchive.org/release/{mbid}/front" -o blue.jpg
curl -s "https://coverartarchive.org/release/{mbid}/front" -o court-and-spark.jpg
# Download artist images from fanart.tv
cd ../artists
curl -s "https://fanart.tv/..." -o joni-mitchell-1970.jpg
# Extract video thumbnails
cd ../live
ffmpeg -i ../../video/isle-of-wight-1970.mp4 -ss 00:01:30 -frames:v 1 -q:v 2 isle-of-wight-1970.jpg
# Reference artwork by path
for file in music/Joni\ Mitchell/Blue/*.opus; do
opustags -i "$file" --set-cover artwork/albums/blue.jpg
done
# Or create symlink in album directory
ln -s ../../artwork/albums/blue.jpg music/Joni\ Mitchell/Blue/cover.jpg
Record source and rights for all artwork:
# Artwork Attribution
## Albums
### blue.jpg
- Source: MusicBrainz Cover Art Archive
- URL: https://coverartarchive.org/release/5c004fe3-6e96-3f43-9a85-16e5be8805fa
- License: Public Domain
- Uploaded by: user123
- Date: 2015-03-20
### court-and-spark.jpg
- Source: fanart.tv
- URL: https://fanart.tv/music/...
- License: Fair Use
- Date: 2018-07-15
## Artists
### joni-mitchell-1970.jpg
- Source: Extracted from "Live at Isle of Wight 1970" video
- License: Fair Use
- Date: 2026-02-14
## Live
### isle-of-wight-1970.jpg
- Source: Video frame extraction (ffmpeg)
- Original video: Live at Isle of Wight 1970.mp4
- License: Fair Use
- Date: 2026-02-14
# opustags does not support extraction, use ffmpeg
ffmpeg -i input.opus -an -vcodec copy cover.jpg
# Extract first attached picture
ffmpeg -i input.mp4 -map 0:v -map -0:V -c copy cover.jpg
# Extract all video streams (including artwork)
ffmpeg -i input.mp4 -map 0:v -c copy "stream-%d.jpg"
# Extract cover art
ffmpeg -i input.mp3 -an -vcodec copy cover.jpg
# Using id3v2 (alternative)
id3v2 -l input.mp3 # List frames
id3v2 --APIC input.mp3 > cover.jpg # Extract
# Extract using metaflac
metaflac --export-picture-to=cover.jpg input.flac
# Or ffmpeg
ffmpeg -i input.flac -an -vcodec copy cover.jpg
# Opus
opustags input.opus | grep -i picture
# MP4
ffprobe -v error -select_streams v -show_entries stream=disposition:stream_tags input.mp4
# MP3
ffprobe -v error -show_entries format_tags input.mp3 | grep -i picture
# Or extract and check file size
ffmpeg -i input.mp3 -an -vcodec copy test.jpg
ls -lh test.jpg # Should be >50KB for quality cover
# Using identify (ImageMagick)
identify cover.jpg
# Using ffprobe
ffprobe -v error -select_streams v:0 -show_entries stream=width,height cover.jpg
# Check cover file size
ls -lh cover.jpg
# Ideal: 100KB - 500KB
# Too small (<50KB): low quality
# Too large (>1MB): unnecessary bloat
# Check if cover is actually embedded
ffprobe -v error -show_entries format_tags input.mp3
# Re-embed with correct disposition
ffmpeg -i input.mp4 -i cover.jpg -map 0 -map 1 -c copy \
-disposition:v:1 attached_pic \
output.mp4
# Convert to JPEG
convert input.png output.jpg
# Re-embed converted image
opustags -i input.opus --set-cover output.jpg
# Resize and optimize
convert cover.jpg -resize 1200x1200 -quality 85 cover-optimized.jpg
# Re-embed optimized image
opustags -i input.opus --set-cover cover-optimized.jpg
# Try fanart.tv
curl -s "https://webservice.fanart.tv/v3/music/${mbid}?api_key=${FANART_API_KEY}" | \
jq -r '.albums."'"$album_mbid"'".albumcover[0].url'
# Or Google Images (manual)
# Or extract from video if available
#!/bin/bash
# Download cover from MusicBrainz, process, and embed
file="$1"
# Get album MBID from existing tags
album=$(opustags "$file" | grep "ALBUM=" | cut -d= -f2)
artist=$(opustags "$file" | grep "ARTIST=" | cut -d= -f2)
# Query MusicBrainz
artist_enc=$(echo "$artist" | jq -sRr @uri)
album_enc=$(echo "$album" | jq -sRr @uri)
mbdata=$(curl -s "https://musicbrainz.org/ws/2/release/?query=artist:${artist_enc}%20AND%20release:${album_enc}&fmt=json" \
-H "User-Agent: MediaCurator/1.0")
mbid=$(echo "$mbdata" | jq -r '.releases[0].id')
# Download cover
cover_url="https://coverartarchive.org/release/${mbid}/front"
curl -s "$cover_url" -o "/tmp/cover-${mbid}.jpg"
# Process: resize, optimize, strip metadata
convert "/tmp/cover-${mbid}.jpg" \
-resize 1200x1200 \
-quality 85 \
-strip \
"/tmp/cover-${mbid}-processed.jpg"
# Embed into file
opustags -i "$file" --set-cover "/tmp/cover-${mbid}-processed.jpg"
# Save to canonical directory
artist_safe=$(echo "$artist" | tr '/' '-')
album_safe=$(echo "$album" | tr '/' '-')
mkdir -p "artwork/albums"
cp "/tmp/cover-${mbid}-processed.jpg" "artwork/albums/${artist_safe}-${album_safe}.jpg"
# Update ATTRIBUTION.md
cat >> artwork/ATTRIBUTION.md <<EOF
### ${artist_safe}-${album_safe}.jpg
- Source: MusicBrainz Cover Art Archive
- URL: ${cover_url}
- MBID: ${mbid}
- Date: $(date +%Y-%m-%d)
EOF
echo "Cover embedded: $file"
echo "Saved to: artwork/albums/${artist_safe}-${album_safe}.jpg"
/tag-collection@metadata-tagging@Metadata Curatordata-ai
Report which research-corpus radar sidecars are overdue for refresh. Computes staleness (days since last refresh vs the cadence window) for every radar, sorted most-overdue-first. Runs via `aiwg corpus radar-status`.
data-ai
Aggregate research-corpus radar sidecars into a corpus or per-cluster freshness report — totals, overdue count, per-cluster / per-GRADE / per-trajectory breakdowns, an overdue table, and per-radar rationale snippets. Runs via `aiwg corpus radar-report`.
testing
Scaffold radar/freshness sidecars for research-corpus REFs. Pulls title/authors from the citation sidecar and GRADE from the analysis doc, defaults the refresh cadence from GRADE and the cluster from a corpus-local map, and stamps documentation/radar/REF-XXX-radar.md. Runs via `aiwg corpus radar-init`.
data-ai
Compute an entity's publication trajectory — per-year paper counts, topic drift, hot-streak detection (≥3 consecutive A-grade years), and career phase. Runs via `aiwg corpus profile-temporal`.