IG Reel Extract — Download & Analyze Social Reels

Download, frame-capture, and transcribe Instagram Reels for competitor research, script analysis, and content repurposing.

When to Use

• Competitor research: pull scripts from top-performing competitor videos
• Hook analysis: study what hooks work in your niche
• Content repurposing: extract and adapt transcripts
• Voice profiling: feed to toneprint or voice-style-extractor skill

Proxy Rule — Mandatory

Instagram downloads MUST use Oxylabs proxy. No home IP scraping.

Read proxy from D:\Ecosystem\secrets\MASTER_API_KEYS.env (grep for PROXY_1). Format: host:port:user:pass → yt-dlp proxy format: http://user:pass@host:port

Execution Steps

Step 1: Create output directory

mkdir D:/ClaudeDev/reel-extracts/<slug>

Use reel ID or timestamp as slug (e.g., reel-2026-05-12-1830).

Step 2: Run extractor with proxy

python "C:\Users\mikem\.claude\skills\ig-reel-extract\scripts\extract_reel.py" \
  "<URL>" \
  --out "D:/ClaudeDev/reel-extracts/<slug>" \
  --proxy "http://user:pass@host:port"

Step 3: Read results

Output files:

• metadata.json — title, author, duration, view count, like count
• transcript.txt — full transcript if captions available
• frame_*.jpg — key frames captured at intervals

Step 4: Analyze

Present:

• Video downloaded: yes/no + file size
• Frames captured: list with timestamps
• Transcript: full text
• Hook analysis: first 3 seconds script
• CTA: what the video asks viewers to do

Output Folder Structure

D:/ClaudeDev/reel-extracts/
  reel-2026-05-12-1830/
    metadata.json
    transcript.txt
    video.mp4
    frame_0s.jpg
    frame_5s.jpg
    frame_10s.jpg

Python Pattern (Direct)

import subprocess, json, os

def extract_reel(url, out_dir, proxy):
    result = subprocess.run([
        "python",
        r"C:\Users\mikem\.claude\skills\ig-reel-extract\scripts\extract_reel.py",
        url,
        "--out", out_dir,
        "--proxy", proxy
    ], capture_output=True, text=True)
    
    metadata_path = os.path.join(out_dir, "metadata.json")
    with open(metadata_path) as f:
        return json.load(f)

Hook Analysis Framework

After extraction, analyze the transcript's first 3 seconds:

| Element | Question | |---------|---------| | Pattern interrupt | Does it break expected format? | | Identity call-out | Does it name a specific audience? | | Curiosity gap | Does it create immediate question? | | Bold claim | Is there a specific, surprising claim? | | Visual hook | What happens on screen in second 1? |

Related Topics

• [[social-transcript-extractor]] — extract transcripts without downloading (API method)
• [[ytdlp-agent]] — YouTube-specific download and transcript CLI
• [[scriptwriter]] — use extracted hooks as reference for new scripts
• [[ai-video-pipeline]] — produce competitor-informed videos

#video-sop #instagram #reel #extract #competitor-research #transcript