npx skills add ...
npx skills add curiositech/some_claude_skills --skill video-processing-editing
FFmpeg automation for cutting, trimming, concatenating videos. Audio mixing, timeline editing, transitions, effects. Export optimization for YouTube, social media. Subtitle handling, color
npx skills add curiositech/some_claude_skills --skill video-processing-editing
Expert in FFmpeg-based video editing, processing automation, and export optimization for modern content creation workflows.
✅ Use for:
❌ NOT for:
| Tool | Speed | Features | Use Case |
|---|---|---|---|
| FFmpeg | Very Fast | CLI automation | Production pipelines |
| MoviePy | Medium | Python API | Programmatic editing |
| PyAV | Fast | Low-level control | Custom processing |
| DaVinci Resolve | Slow | Full NLE | Manual editing |
Decision tree:
Novice thinking: "Just cut the video at any timestamp"
Problem: Causes artifacts, black frames, and playback issues.
Wrong approach:
Why wrong:
-c copy (stream copy) without keyframe alignment breaks playbackCorrect approach 1: Re-encode for precise cuts
Correct approach 2: Keyframe-aligned stream copy
Correct approach 3: Two-pass for best of both worlds
Performance comparison:
| Method | Time (1-hour video) | Accuracy | Quality |
|---|---|---|---|
| Stream copy (arbitrary) | 2s | ❌ Broken | ❌ Artifacts |
| Stream copy (keyframe) | 2s | ±2s | ✅ Perfect |
| Re-encode (simple) | 15min | ✅ Frame | ⚠️ Quality loss |
| Two-pass (optimal) | 3min | ✅ Frame | ✅ Perfect |
Timeline context:
-c copy added for stream copyingNovice thinking: "Apply all edits in one FFmpeg command"
Problem: Multiple re-encodings cause cumulative quality loss.
Wrong approach:
Why wrong:
Correct approach 1: Chain operations in single command
Correct approach 2: Use stream copy when possible
Quality comparison:
| Method | Encoding Passes | Quality (VMAF) | Time |
|---|---|---|---|
| 3x re-encode (CRF 23) | 3 | 82/100 | 45min |
| Single pass (CRF 23) | 1 | 91/100 | 15min |
| Stream copy + 1 encode | 1 | 95/100 | 18min |
| All stream copy | 0 | 100/100 | 30s |
Novice thinking: "Just concatenate videos together"
Problem: Color shifts, mismatched brightness, broken playback.
Wrong approach:
Why wrong:
Correct approach:
Color space guide:
| Standard | Color Space | Transfer | Primaries | Use Case |
|---|---|---|---|---|
| BT.601 | SD | bt470bg | bt470bg | Old SD content |
| BT.709 | HD | bt709 | bt709 | Modern HD/FHD |
| BT.2020 | UHD/HDR | smpte2084 | bt2020 | 4K HDR |
| sRGB | Web | iec61966-2-1 | bt709 | Web delivery |
Novice thinking: "Video and audio are separate, just overlay them"
Problem: Lip sync issues, audio drift, broken playback.
Wrong approach:
Why wrong:
Correct approach 1: Stretch/compress audio to match video
Correct approach 2: Precise offset and trim
Correct approach 3: Mix multiple audio tracks with sync
Audio sync checklist:
Novice thinking: "One export settings for everything"
Problem: Wasted bandwidth, poor quality, rejected uploads, compatibility issues.
Wrong approach:
Why wrong:
Correct approach: Platform-optimized exports
YouTube (recommended settings):
Instagram (Stories, Reels, Feed):
Twitter/X:
TikTok:
Web (HTML5 video):
Platform specs table:
| Platform | Max Size | Max Duration | Resolution | FPS | Bitrate | Codec |
|---|---|---|---|---|---|---|
| YouTube | Unlimited | Unlimited | 8K | 60 | Auto | H.264/VP9 |
| Instagram Story | 100 MB | 15s | 1080x1920 | 30 | ~5 Mbps | H.264 |
| Instagram Reel | 1 GB | 90s | 1080x1920 | 30 | ~8 Mbps | H.264 |
| 512 MB | 2:20 | 1920x1080 | 60 | 5 Mbps | H.264 | |
| TikTok | 287 MB | 10min | 1080x1920 | 30 | ~4 Mbps | H.264 |
| 5 GB | 10min | 1920x1080 | 30 | 5 Mbps | H.264 | |
| Web | Varies | Varies | 1920x1080 | 30 | 2-5 Mbps | H.264 |
Export optimization checklist:
| Scenario | Appropriate? |
|---|---|
| Automated video pipeline (script → video) | ✅ Yes - FFmpeg automation |
| Batch process 100 videos | ✅ Yes - parallel FFmpeg jobs |
| Trim/cut clips programmatically | ✅ Yes - precise cutting |
| Add subtitles to videos | ✅ Yes - burn or soft subs |
| Color grade footage | ⚠️ Limited - basic only |
| Multi-cam editing | ❌ No - use DaVinci Resolve |
| Motion graphics | ❌ No - use After Effects |
| Real-time preview editing | ❌ No - use Premiere/Resolve |
/references/ffmpeg-guide.md - Complete FFmpeg command reference/references/timeline-editing.md - Timeline concepts, multi-track editing/references/export-optimization.md - Platform-specific export settingsscripts/video_editor.py - Cut, trim, concatenate, transitions, effectsscripts/batch_processor.py - Parallel batch video processingThis skill guides: Video editing | FFmpeg | Timeline editing | Transitions | Export optimization | Audio mixing | Color grading | Automated video production
# ✅ Re-encode for frame-accurate cutting
ffmpeg -i input.mp4 -ss 00:01:23.456 -to 00:02:45.678 \
-c:v libx264 -crf 18 -preset medium \
-c:a aac -b:a 192k \
output.mp4
# Frame-accurate, but slower (re-encoding)# ✅ Fast cutting with keyframe alignment
# Step 1: Find keyframes near cut points
ffprobe -select_streams v -show_frames -show_entries frame=pkt_pts_time,key_frame \
-of csv input.mp4 | grep ",1$" | awk -F',' '{print $2}'
# Step 2: Cut at nearest keyframes (fast, no re-encoding)
ffmpeg -i input.mp4 -ss 00:01:22.000 -to 00:02:46.000 -c copy output.mp4
# Blazing fast, no quality loss, but not frame-accurate# ✅ Fast seek + precise cut
ffmpeg -ss 00:01:20.000 -i input.mp4 \
-ss 00:00:03.456 -to 00:01:25.678 \
-c:v libx264 -crf 18 -preset medium \
-c:a aac -b:a 192k \
output.mp4
# -ss BEFORE -i: Fast seek to keyframe (no decode)
# -ss AFTER -i: Precise trim (only decode needed portion)# ❌ Re-encode for each operation (quality degradation)
# Operation 1: Trim
ffmpeg -i input.mp4 -ss 00:01:00 -to 00:05:00 \
-c:v libx264 -crf 23 temp1.mp4
# Operation 2: Add audio
ffmpeg -i temp1.mp4 -i audio.mp3 -c:v libx264 -crf 23 \
-map 0:v -map 1:a temp2.mp4
# Operation 3: Add subtitles
ffmpeg -i temp2.mp4 -vf subtitles=subs.srt \
-c:v libx264 -crf 23 output.mp4
# Result: 3x re-encoding = significant quality loss# ✅ Single-pass encoding with all operations
ffmpeg -ss 00:01:00 -i input.mp4 -i audio.mp3 \
-to 00:04:00 \
-vf "subtitles=subs.srt" \
-map 0:v -map 1:a \
-c:v libx264 -crf 18 -preset medium \
-c:a aac -b:a 192k \
output.mp4
# Single re-encode, all operations applied at once# ✅ Lossless operations with stream copy
# Trim (stream copy)
ffmpeg -i input.mp4 -ss 00:01:00 -to 00:05:00 -c copy temp.mp4
# Add audio (stream copy video, encode audio)
ffmpeg -i temp.mp4 -i audio.mp3 \
-map 0:v -map 1:a \
-c:v copy -c:a aac -b:a 192k \
temp2.mp4
# Burn subtitles (must re-encode video)
ffmpeg -i temp2.mp4 -vf subtitles=subs.srt \
-c:v libx264 -crf 18 -preset medium \
-c:a copy \
output.mp4
# Only 1 video re-encode (for subtitles)# ❌ Concatenate videos with different color spaces
# clip1.mp4: BT.709 (HD), yuv420p
# clip2.mp4: BT.601 (SD), yuvj420p (full range)
# clip3.mp4: BT.2020 (HDR), yuv420p10le
# Create concat list
echo "file 'clip1.mp4'" > list.txt
echo "file 'clip2.mp4'" >> list.txt
echo "file 'clip3.mp4'" >> list.txt
# Concatenate without color normalization
ffmpeg -f concat -safe 0 -i list.txt -c copy output.mp4
# Result: Color shifts between clips, broken HDR metadata# ✅ Normalize color space before concatenation
# Step 1: Analyze color space of each clip
ffprobe -v error -select_streams v:0 \
-show_entries stream=color_space,color_transfer,color_primaries,pix_fmt \
-of default=noprint_wrappers=1 clip1.mp4
# Step 2: Normalize all clips to common color space
# Target: BT.709 (HD), yuv420p, limited range
# Normalize clip1 (already BT.709)
ffmpeg -i clip1.mp4 -c copy clip1_normalized.mp4
# Normalize clip2 (BT.601 SD → BT.709 HD)
ffmpeg -i clip2.mp4 \
-vf "scale=in_range=full:out_range=limited,colorspace=bt709:iall=bt601:fast=1" \
-color_primaries bt709 \
-color_trc bt709 \
-colorspace bt709 \
-c:v libx264 -crf 18 -preset medium \
-c:a copy \
clip2_normalized.mp4
# Normalize clip3 (BT.2020 HDR → BT.709 SDR)
ffmpeg -i clip3.mp4 \
-vf "zscale=t=linear:npl=100,format=gbrpf32le,zscale=p=bt709,tonemap=hable:desat=0,zscale=t=bt709:m=bt709:r=limited,format=yuv420p" \
-color_primaries bt709 \
-color_trc bt709 \
-colorspace bt709 \
-c:v libx264 -crf 18 -preset medium \
-c:a copy \
clip3_normalized.mp4
# Step 3: Concatenate normalized clips
echo "file 'clip1_normalized.mp4'" > list.txt
echo "file 'clip2_normalized.mp4'" >> list.txt
echo "file 'clip3_normalized.mp4'" >> list.txt
ffmpeg -f concat -safe 0 -i list.txt -c copy output.mp4# ❌ Replace audio without sync consideration
ffmpeg -i video.mp4 -i audio.mp3 \
-map 0:v -map 1:a \
-c:v copy -c:a copy \
output.mp4
# Problems:
# - Audio duration ≠ video duration
# - No audio stretching/compression
# - Drift over time# ✅ Adjust audio speed to match video duration
# Get durations
VIDEO_DUR=$(ffprobe -v error -show_entries format=duration \
-of default=noprint_wrappers=1:nokey=1 video.mp4)
AUDIO_DUR=$(ffprobe -v error -show_entries format=duration \
-of default=noprint_wrappers=1:nokey=1 audio.mp3)
# Calculate speed ratio
RATIO=$(echo "$VIDEO_DUR / $AUDIO_DUR" | bc -l)
# Stretch audio to match video (with pitch correction)
ffmpeg -i video.mp4 -i audio.mp3 \
-filter_complex "[1:a]atempo=${RATIO}[a]" \
-map 0:v -map "[a]" \
-c:v copy -c:a aac -b:a 192k \
output.mp4# ✅ Sync audio with offset and trim
# Audio starts 0.5s late, trim to match video
ffmpeg -i video.mp4 -itsoffset 0.5 -i audio.mp3 \
-map 0:v -map 1:a \
-shortest \
-c:v copy -c:a aac -b:a 192k \
output.mp4
# -itsoffset: Delay audio by 0.5s
# -shortest: Trim to shortest stream# ✅ Mix dialogue, music, effects with precise timing
ffmpeg -i video.mp4 -i dialogue.wav -i music.mp3 -i sfx.wav \
-filter_complex "
[1:a]adelay=0|0[dlg];
[2:a]volume=0.3,adelay=500|500[mus];
[3:a]adelay=1200|1200[sfx];
[dlg][mus][sfx]amix=inputs=3:duration=first[a]
" \
-map 0:v -map "[a]" \
-c:v copy -c:a aac -b:a 256k \
output.mp4
# adelay: Precise millisecond timing
# amix: Mix multiple audio streams
# volume: Normalize levels□ Verify video and audio durations match
□ Use -shortest to prevent excess audio
□ Apply adelay for precise timing offsets
□ Use atempo for speed adjustment (maintains pitch)
□ Set audio bitrate appropriately (128k-256k)
□ Test lip sync at beginning, middle, end# ❌ Export everything at 4K 50 Mbps
ffmpeg -i input.mp4 \
-c:v libx264 -b:v 50M -s 3840x2160 \
-c:a aac -b:a 320k \
output.mp4
# For Instagram story: 2 GB file, rejected (max 100 MB)
# For YouTube: Could use 10 Mbps and look identical
# For Twitter: Exceeds bitrate limits# ✅ YouTube 1080p upload
ffmpeg -i input.mp4 \
-c:v libx264 -preset slow -crf 18 \
-s 1920x1080 -r 30 \
-pix_fmt yuv420p \
-color_primaries bt709 -color_trc bt709 -colorspace bt709 \
-movflags +faststart \
-c:a aac -b:a 192k -ar 48000 \
youtube_1080p.mp4
# YouTube 4K upload
ffmpeg -i input.mp4 \
-c:v libx264 -preset slow -crf 18 \
-s 3840x2160 -r 60 \
-pix_fmt yuv420p \
-movflags +faststart \
-c:a aac -b:a 256k -ar 48000 \
youtube_4k.mp4# ✅ Instagram Story (9:16, max 100 MB, 15s)
ffmpeg -i input.mp4 \
-c:v libx264 -preset medium -crf 23 \
-s 1080x1920 -r 30 -t 15 \
-pix_fmt yuv420p \
-movflags +faststart \
-c:a aac -b:a 128k \
instagram_story.mp4
# ✅ Instagram Reel (9:16, max 90s)
ffmpeg -i input.mp4 \
-c:v libx264 -preset medium -crf 23 \
-s 1080x1920 -r 30 -t 90 \
-pix_fmt yuv420p \
-movflags +faststart \
-c:a aac -b:a 128k \
instagram_reel.mp4
# ✅ Instagram Feed (1:1 or 4:5)
ffmpeg -i input.mp4 \
-c:v libx264 -preset medium -crf 23 \
-s 1080x1080 -r 30 \
-pix_fmt yuv420p \
-movflags +faststart \
-c:a aac -b:a 128k \
instagram_feed.mp4# ✅ Twitter video (max 512 MB, 2:20)
ffmpeg -i input.mp4 \
-c:v libx264 -preset medium -crf 23 \
-s 1280x720 -r 30 -t 140 \
-maxrate 5000k -bufsize 10000k \
-pix_fmt yuv420p \
-movflags +faststart \
-c:a aac -b:a 128k \
twitter.mp4# ✅ TikTok (9:16, max 287 MB, 10 min)
ffmpeg -i input.mp4 \
-c:v libx264 -preset medium -crf 23 \
-s 1080x1920 -r 30 -t 600 \
-pix_fmt yuv420p \
-movflags +faststart \
-c:a aac -b:a 128k \
tiktok.mp4# ✅ Web optimized (fast load, broad compatibility)
ffmpeg -i input.mp4 \
-c:v libx264 -preset medium -crf 23 \
-s 1920x1080 -r 30 \
-pix_fmt yuv420p \
-profile:v baseline -level 3.0 \
-movflags +faststart \
-c:a aac -b:a 128k -ar 48000 \
web.mp4□ Use -movflags +faststart for web (progressive download)
□ Use -pix_fmt yuv420p for broad compatibility
□ Set -r 30 for most platforms (avoid variable framerate)
□ Use -preset slow for final exports (better quality)
□ Use -preset ultrafast for drafts
□ Apply -maxrate and -bufsize for streaming
□ Test playback on target platform before bulk export□ Align cuts to keyframes (or two-pass seek)
□ Chain operations in single FFmpeg command
□ Normalize color spaces before concatenating
□ Verify audio/video sync (test at multiple points)
□ Use platform-specific export presets
□ Apply -movflags +faststart for web delivery
□ Set proper color metadata (bt709 for HD)
□ Test output file on target platform
□ Keep lossless intermediate files (ProRes, FFV1)
□ Use hardware acceleration for batch jobs (NVENC, VideoToolbox)