Turns real-time YouTube trend data into SEO-scored titles, full video concepts and structured scripts, with a content engine built never to invent statistics.
- Built at
- RaftWorks
- Client
- Scriva
- Role
- Full-stack developer
- Platform
- Web app
- Year
- 2025
Creators guess at what to make next. Trend data exists, but turning it into a title, a concept and a script is hours of work, and generic writing tools happily invent statistics.
- Integrated six YouTube Data API v3 endpoints for trend discovery
- Built the content engine: eight purpose-built prompts and a JSON normalisation workflow
- Built credit-based billing with Stripe and Gumroad
- Implemented the product UI from the team's Figma designs
Six YouTube Data API calls power discovery: trending videos by region, keyword search by view count, batch video details, and category, language and region metadata, surfaced in a Trends & Analytics dashboard with a keyword explorer.
Rather than one generic generation call, the engine runs eight specialised prompts. Several enforce a strict never-fabricate rule: every statistic or URL must come from a real web search, or the answer says no current data is available. Others are barred from statistics entirely, forcing story-driven framing.
Every response passes through a backend JSON-normalisation workflow that guarantees its shape before it reaches the UI. That's the mechanism behind the Clickability, SEO and Retention scores on every title.
Plans and credit packs through Stripe and Gumroad, metered per generation.
- Platform
- Bubble.ioBackend workflows
- Integrations
- YouTube Data API v3OpenAI APIStripeGumroad
Built end to end: trend discovery, a constrained content engine and credit billing.
- YouTube API endpoints
- 6
- purpose-built prompts
- 8
- billing providers
- 2
Scroll the grid · click any screen to open it larger · ← → to browse