Moonlit Showrunner v0.7.3 — Kitten-Centered Character-Locking Prototype This release marks the current working prototype of Moonlit Showrunner, an AI-assisted short-film pipeline built with Python and Streamlit. Starting from a story premise, the app generates a structured story package, character profiles, script, storyboard, visual prompts, continuity notes, character reference cards, generated scene images, a visual animatic MP4, and optional Sora video clips assembled into a final video output. Version 0.7.3 focuses on the kitten-centered version of the project, following Milo, a small kitten who loses his first tooth and discovers a moonlit fairy archive where tiny keepsakes become stars. Highlights Kitten-centered default story premise featuring Milo and the Tooth Fairy Character continuity profiles for more stable visual identity Reusable character reference-card generation Continuity-aware scene image prompts Visual animatic MP4 assembly using generated still images Optional Sora video generation, one clip per storyboard shot Concatenation of Sora clips into a full MP4 output Moderation-aware Sora prompt sanitization Graceful handling for moderation and billing-limit errors Updated README with current screenshots and generated video link Generated video output The current Sora-generated video output is available here: https://youtu.be/vknl0erAFn8 Notes This is a working prototype, not a finished production animation tool. The project explores how generative AI can support creative production through structured, reviewable, human-in-the-loop stages rather than relying on a single prompt-to-video workflow. Known limitations include imperfect character consistency, occasional visual drift, long-running generation steps, and the need for stronger resume support after Streamlit restarts. Next planned improvements Resume support for loading an existing run folder Single-shot regeneration controls Better reuse of existing character reference cards Prompt editing before image or video generation Continuity scoring and run comparison Improved review workflow for generated assets