shorts-video-maker

Shorts Video Maker Skill

Extracts highlight segments from the original YouTube video and creates 9:16 vertical Shorts clips.

Requirements

System Dependencies

• Python 3.8+
• yt-dlp (YouTube download)
• ffmpeg (video editing)
• OpenAI Whisper (speech recognition) or Whisper API

Installation

pip install yt-dlp openai-whisper
brew install ffmpeg  # macOS
# 또는 apt-get install ffmpeg  # Ubuntu

Workflow

Step 1: Download Video

python scripts/download_video.py --url "YOUTUBE_URL" --output "temp/"

• Download the highest quality video with yt-dlp
• Output: temp/video.mp4

Step 2: Extract Subtitles

python scripts/transcribe.py --input "temp/video.mp4" --output "temp/transcript.json"

• Convert speech to text with Whisper
• Output JSON with timestamps
• Output: temp/transcript.json

Step 3: Reference Shorts Script

• outputs/shorts-scripts/shorts-01.md ~ shorts-03.md Read
• Check segment information in the "Original Video Reference" section of the script
• Or automatically search for matching segments in transcript.json

Step 4: Extract and Edit Clips

python scripts/cut_shorts.py \
  --input "temp/video.mp4" \
  --transcript "temp/transcript.json" \
  --scripts "outputs/shorts-scripts/" \
  --output "outputs/shorts-videos/"

• Clip segments with ffmpeg
• Crop to 9:16 ratio
• Hardcode subtitles (optional)
• Output: outputs/shorts-videos/shorts-01.mp4 ~ shorts-03.mp4

Shorts Specs

Video Specifications

• Resolution: 1080x1920px (vertical)
• Aspect Ratio: 9:16
• Length: 15-60초
• Format: MP4 (H.264)

Subtitle Style

• Position: bottom center or top
• Font: bold sans-serif
• Color: white + black outline
• Size: 80% of screen width

Script Details

download_video.py

• Enter YouTube URL
• Download highest quality
• Show progress
• Error handling

transcribe.py

• Extract audio
• Load Whisper model
• Generate timestamps per segment
• Output in JSON format

cut_shorts.py

• Clip based on start/end time
• Convert 16:9 → 9:16 (center crop)
• Subtitle overlay (optional)
• Batch process 3 clips

Output

Final result

• outputs/shorts-videos/shorts-01.mp4
• outputs/shorts-videos/shorts-02.mp4
• outputs/shorts-videos/shorts-03.mp4

Intermediate files (temp/)

• temp/video.mp4: original download
• temp/transcript.json: subtitle data
• temp/audio.wav: extracted audio

Reference Files

• references/shorts-spec.md: Shorts platform-specific specs
• ../shorts-scriptwriter.md: script agent (integration)

Usage Examples

# 전체 파이프라인 실행
cd .claude/skills/shorts-video-maker

# 1. 다운로드
python scripts/download_video.py \
  --url "https://youtube.com/watch?v=xxxxx" \
  --output "temp/"

# 2. 자막 추출
python scripts/transcribe.py \
  --input "temp/video.mp4" \
  --output "temp/transcript.json"

# 3. 쇼츠 생성
python scripts/cut_shorts.py \
  --input "temp/video.mp4" \
  --transcript "temp/transcript.json" \
  --scripts "outputs/shorts-scripts/" \
  --output "outputs/shorts-videos/"

Notes

1. YouTube video copyright check required
2. Crop adjustment required if the original video is not 16:9
3. Accuracy/speed tradeoff depending on Whisper model size
4. Processing speed improves when using a GPU