It all started from a real practical need: I needed a fast way to translate scripts into clear shots before arriving on set, without having to hand-draw every single panel. This sparked the creation of StoryAid, a platform for generating paginated, ready-to-use storyboards from Fountain/Final Draft scripts, free text descriptions, or simple sketches.

The core of the system is a flexible AI engine: it can run entirely locally using Ollama and ComfyUI (optimized for 12GB GPUs), or rely on the cloud via Google Gemini models. To maintain visual consistency across different scenes, I implemented a "visual bible" (characters, locations, props) that guides the model. This is paired with a vector camera symbology system (pan, tilt, zoom, dolly arrows) that always prints perfectly sharp.

The pipeline is completed by automatic A4 pagination, version tracking for each panel, and direct export to animatic or CSV shot lists, all easily deployable via Docker Compose.

And what if in the future we could generate entire video pre-visualizations straight from a simple text file?