# reusable character library for consistent AI video — spoolcast

A reusable character and prop library — the World Kit — so a person or object looks the same in every shot of a video, and in every video of a series.

Type: project

Published: 2026-07-26

[Original page](<https://artlu.ai/project/reusable-character-library-for-consistent-ai-video-spoolcast>)

[Browse this group](<https://artlu.ai/ai/groups/spoolcast.md>)

Parent project: spoolcast

Series: features

Recorded status: launched

Updated: 2026-09-16T17:00:34.585Z

Technologies: Python, kie.ai, React, TypeScript, Cloudflare R2

## Details

**What is it?**
The part of spoolcast that fixes the flaw everyone notices in AI video: the character whose face changes between shots.

**What it does**
- A shared, read-only character library with reusable variants of each character
- Characters, props, and places inherit downward — global library, then template, then series, then the individual video
- Every place gets a fixed time of day, so lighting and mood can't drift from shot to shot
- Image prompts automatically carry the saved description of each character and object in the frame
- A check refuses any script that puts a character on screen without their look rules

**How it works**
Consistency comes from reference images attached at generation time, not from describing a face in words and hoping. A character's look lives in one record that every shot reads, so correcting it once corrects it everywhere. Shots list every character on screen, which stopped the model from drawing a stranger when someone was mentioned but not listed.

**Built with**
Python, kie.ai for image generation, Cloudflare R2 for reference storage, React + TypeScript front end.

## Links

[Site or demo](<https://spoolcast-web.pages.dev/>)


Status and dates describe the published record, not a recent test of an external service. Missing details are not inferred. Public links do not imply an open-source licence or permission to reuse material.

[Directory](<https://artlu.ai/llms.txt>)
