Repurpose AI
20 days
1. The Founder's Problem
Content creators spend 6+ hours manually adapting one video transcript into platform-specific posts. Each platform has different formats, character limits, and conventions. Manual repurposing is slow, inconsistent, and kills content velocity.
2. Why Traditional Solutions Failed
- ✗Generic AI tools don't understand platform-specific conventions (LinkedIn vs Twitter structure)
- ✗Manual writing takes hours per platform and doesn't scale
- ✗Copy-pasting transcripts produces generic content that doesn't engage audiences
3. What We Built
- ✓YouTube transcript extraction via URL (auto-fetches captions if available)
- ✓6 platform-specific output formats: LinkedIn long-form, LinkedIn hooks, Twitter threads, email newsletters, YouTube descriptions, short-form video scripts
- ✓Groq AI with Llama 3.3 70B for content generation with platform-aware prompts
- ✓Tone control: Professional, Casual, Punchy with regeneration on demand
- ✓Project dashboard with full history and usage tracking
- ✓Supabase auth with Google OAuth and tiered pricing (Free: 3 projects/month, Pro: unlimited)
4. The Process
WEEK 1
YouTube API integration (3 days), Transcript parser (2 days), Auth + database schema (2 days)
WEEK 2
AI prompt engineering for 6 formats (4 days), Tone control system (2 days), Dashboard UI (1 day)
WEEK 3
Regeneration logic (2 days), Usage tracking + limits (2 days), Stripe integration + deploy (3 days)
5. Results & Metrics
users
500+ specs generated
formats
6 output formats per transcript
performance
Drafts generated in seconds
plans
Free + Pro ($19/mo planned)
code
React 18, TypeScript, Vite, Supabase
6. Tech Stack & Why
React 18 + Vite
Fast dev experience, instant HMR, lighter than Next.js for this use case
Groq Llama 3.3 70B
800 tokens/s generation speed, $0.59/1M tokens, perfect for content generation
Supabase
Auth + PostgreSQL + Edge Functions in one platform, generous free tier
Tailwind + shadcn/ui
Rapid UI development with consistent design system
7. What We Cut to Ship Fast
- Audio/video file transcription (YouTube URLs cover 80% of use cases)
- Real-time collaboration (single-user MVP is sufficient)
- Custom AI model training (pre-built prompts work well enough)
8. Lessons Learned
- →Platform-specific prompts are 60% of the value. Generic AI output doesn't work.
- →Users want drafts, not final copy. Regeneration is more important than perfection.
- →YouTube transcript extraction is harder than expected many videos lack captions.
Ready to Build Your MVP?
Let's create something similar for your idea. I can build and launch in 2-3 weeks.
Start Your Project