Cinecule

Rendering Complete
Custom API's that help the AI understand how to differentiate between source types.
Link Capabilities
Home Page (Explanation)
Entity Structure
AWS Architecture

I used nostackai as the backend for Cinecule, a cinematic studio that turns source material (documents and other uploads) into narrated, AI-directed videos. What I built on it • Data model without writing a backend. I defined entities like projects, content sources, generated slides, voiceovers, camera shots and export assets. The platform generated the create, read, update, delete and list endpoints for each, so the frontend could talk to them straight away. • A multi-step AI pipeline as Logic Blocks. Ingesting a source, generating a storyboard, writing cinematic direction, producing a voiceover and starting a render are each a Logic Block. They chain AI steps with data steps in one place. • Custom code where I needed it (Glyphs). I used Glyphs to parse the AI's storyboard JSON into database rows, generate ElevenLabs narration for every slide and hand the frontend a download link for the audio. • A GitHub connector for rendering. A Logic Block triggers a GitHub Actions job that renders the video, with no extra service of my own. • Events. Event-triggered runs let ingesting a file start the pipeline automatically.

What worked well • I could go from an idea to a working AI video pipeline quickly because the database, API, auth, file uploads and AI calls were already there. • Mixing no-code steps with small pieces of real code, instead of choosing one or the other, was the best part.

What was hard • Logic Blocks started by an event behaved differently from ones I ran by hand in the console. Some steps (running Glyph code, reading uploaded files) failed only in event runs, and I had to work out that the cause was platform configuration, not my logic. The console's test run is a dry run, so it couldn't show me this. • Plan limits were unclear. I hit a surprise 10 MB upload cap while rendering a 12 MB video, even on a plan that should allow more, and the limit wasn't listed anywhere I could see. Would I use it again? Yes, for building AI-heavy apps quickly where I want a database, API and automation without managing servers.