content-repurposing

name	content-repurposing
description	Content repurposing strategy and extraction rules — how to derive maximum formats from a single generated asset. Triggers on "repurpose", "reuse content", "turn this into", "make a carousel from this video", or when /repurpose command is invoked.

Content Repurposing — One Asset, Many Formats

Repurposing Matrix

Every generated asset contains reusable components. The matrix below maps source formats to viable derivatives:

Source Format	Extractable Assets	Derivative Formats
Reel	transcript, 4 images, voiceover, music, captions	carousel, infographic, blog excerpt, social captions, email snippet, quote card
Ad	transcript, 5 images, voiceover, music, script beats	carousel, infographic, social captions, email snippet, quote card, landing page copy
UGC / UGC Ad	transcript, avatar image, voiceover, B-roll images	testimonial carousel, quote card, social captions, email testimonial
Talking Character	transcript, 4 character images, voiceover	carousel (character slides), social captions, meme-style quote cards
Product Decoded	transcript, assembled/disassembled images, voiceover	infographic, carousel (component breakdown), social captions, spec sheet
Carousel	slide images, caption text	reel script (reverse-engineer), social captions per platform, email content
Infographic	poster image, item cutouts	carousel slides, social posts, ad creative

Asset Extraction Rules

From Video Output Directories

Detect source type from directory naming:

• reel-* → Reel pipeline output
• ad-* → Ad pipeline output
• ugc-* → UGC pipeline output
• ugc-ad-* → UGC Ad pipeline output
• character-* → Talking Character output
• decoded-* → Product Decoded output
• carousel-* → Carousel output
• infographic-* → Infographic output

Transcript Extraction Priority

1. captions.json — Already exists in most video pipelines. Contains word-level timestamps.
2. render-props.json — Contains script text per segment/shot.
3. Whisper transcription — If neither exists, transcribe from voiceover.mp3 or the final .mp4.

Image Extraction

1. From output directory — Use existing images/segment-*.jpg or images/shot-*.jpg
2. From video keyframes — Extract with ffmpeg: ffmpeg -i video.mp4 -vf "select=eq(pict_type\,I)" -vsync vfr frame-%03d.jpg
3. From character images — Use character-shot-*.jpg directly

Derivative Generation Rules

Video → Carousel

1. Extract transcript from captions.json or render-props.json
2. Split transcript into 5-7 key points (one per slide)
3. Use existing segment images as slide backgrounds
4. Generate via /generate-carousel in standard mode with the extracted content
5. Write captions per platform

Slide mapping:

• Slide 1 (Cover): Hook line from transcript (first sentence)
• Slides 2-6: One key insight per slide, sourced from script beats
• Slide 7 (CTA): Brand CTA from config

Video → Infographic

1. Extract key data points from transcript (numbers, comparisons, lists)
2. Use product/item images from output directory
3. Generate via /generate-infographic with extracted data
4. Best when source has comparison or list structure

Video → Social Captions (per platform)

Generate platform-adapted captions from the same transcript:

Platform	Style	Length	Hashtags
Instagram	Story-driven, line breaks, emoji-light	150-300 words	15-20 mixed tiers
TikTok	Casual, hook-first, trending references	50-150 words	3-5 trending
LinkedIn	Professional, insight-led, paragraph format	200-400 words	3-5 industry
Twitter/X	Punchy, thread-ready, single key insight	280 chars or thread	1-2 max
YouTube Shorts	SEO-optimized, keyword-rich description	100-200 words	None (use tags)

Video → Email Snippet

Extract for newsletter or drip campaign use:

• Subject line: Derived from hook (first sentence of script)
• Preview text: Second sentence or support beat
• Body excerpt: 2-3 sentences covering the core insight
• CTA: Link to full video or landing page
• Inline image: First segment image as hero

Video → Quote Card

For static social posts (Instagram Feed, LinkedIn, Twitter):

1. Extract the single most shareable line from transcript
2. Overlay on a branded background (brand colors + font from config)
3. Generate via image service with text overlay prompt
4. Output: 1080x1080 (square) for maximum cross-platform compatibility

Video → Blog Excerpt

For SEO content or newsletter:

1. Expand transcript into 300-500 word article section
2. Add context, sources (from Ideas Bank facts if available)
3. Include the original hook as the H2 heading
4. Embed video link or thumbnail
5. Add internal links to related content

Cost Efficiency

Repurposing costs are minimal compared to original generation:

Derivative	Estimated Cost	Method
Carousel (standard mode)	~$0.00	HTML rendering via Playwright
Carousel (AI character)	~$0.30	New image generation
Social captions (all platforms)	$0.00	LLM text generation only
Email snippet	$0.00	LLM text extraction
Quote card	~$0.02-0.05	Single image generation
Blog excerpt	$0.00	LLM text expansion
Infographic	~$0.10-0.20	Image composition

Total repurpose cost: ~$0.02-0.55 vs original generation cost of $0.24-1.50.

Quality Rules

• Never dilute the hook — the original hook line should remain intact or be strengthened, never weakened
• Platform-native feel — each derivative must feel native to its platform, not like a lazy copy-paste
• Brand consistency — all derivatives use the same brand colors, font, CTA from config
• Fact preservation — if the source contained verified facts, carry the same claims and sources
• No hallucinated content — derivatives can only contain information from the source + brand config, never invented facts