Ficate is building a pipeline that turns chapter fiction into narrated video episodes — shot-by-shot storyboards, generated visuals, continuity-checked subtitles, and automated assembly into finished chapters.
Fully AI-generated video is expensive and inconsistent. We use Claude as the planning and quality-control layer, and generate only what each shot needs.
Claude breaks each chapter into a shot-by-shot storyboard with timing, camera language, and narration cues.
Image and video prompts are written per shot, with locked character and prop references for consistency.
Keyframes and key video clips are generated per shot — no wasted full-length renders.
Generated visuals and subtitles are checked against the source text for continuity errors before assembly.
Automation scripts (FFmpeg-based) assemble images, clips, narration, music, and subtitles into the finished episode.
Our pilot episode — the first chapter of an urban-gothic horror series — was planned, narrated, and quality-controlled with Claude, then assembled end-to-end by the pipeline.

Claude sits at the two places where video pipelines usually break: planning and quality control. It writes the storyboards and prompts, then verifies the generated visuals and subtitles against the source text — so the finished episode actually matches the fiction it came from. Our goal is a full 40+ chapter series at a fraction of the cost of fully AI-generated video.