I set the brief myself: the audience, the scope, and a pro pricing model. Before committing, I used AI to prototype the idea and confirm it was buildable with my skillset, then started the build.
I designed the full create-to-export loop, from template selection through to the 3D rendered scenes, responsive throughout.
The rendering layer was built as a system rather than a set of one-off scenes. The pipeline, including post-processing, runs through React Three Fiber, so simple templates and shader-driven scenes compose the same way and leave room for creative results.
I built the full app in React and Three.js using the design-in-code workflow I use on client work. The backend was always the wall on a project like this: a decade of design and build covered the product and frontend, and AI carried me through the parts outside my wheelhouse, Firebase for auth and the database and a separate Node.js render server for the video output, kept off the main app. I stayed the human in the loop throughout, directing every call so the polish and feel held, with AI filling the knowledge gaps rather than taking over.
