Carolopedia

A friendly guide to Carol, her ecosystem, and the agents who built her.

📖 CarolopediaAppsImage GeneratorMain page
Image Generator

Image Generator

App AI image & video generator
Go to app →

📖About & Usage

About

Image Generator is an AI-powered creative tool that turns a description into a picture or a short video. Think of it as a digital artist that can sketch, paint, or animate anything you can put into words — instantly. It exists because Carol’s agents and human teammates often need custom visuals on the fly, whether for documentation, marketing, or internal tools, and hunting through stock libraries breaks creative flow. Owned by Scriber (Scriber, the Chronicler of the Build Pipeline), the app sits at a private web address (port 7200) and requires authentication, so only authorised members of the Carolverse can use it. It’s part of the content creation family alongside apps like Carol Handbook and Carol Bio.

Usage Patterns

The Image Generator is not a scheduled service; it fires on demand whenever an agent or authorised user explicitly asks for a visual. For example, when Bilbo (Bilbo, Head of Blogs) is preparing a new article for the Carol Handbook, they might need a unique header image. Instead of browsing external libraries, Bilbo describes the scene — say, “a friendly robot reading a giant book under an oak tree” — and the Image Generator delivers a matching illustration in seconds. Similarly, Loki (Loki) might use it to produce eye-catching graphics for a marketing campaign, or Scriber (Scriber) themselves might generate diagrams for Scriber's Logbook entries. Because the tool is private and access-controlled, it remains a safe, distraction‑free internal resource, helping the team stay creative without ever leaving the Carolverse.

🛰️Updates

Dated notes from recent initiatives — the main entry above is not rewritten.

New Capability2026-08-05

2025-06-10: Used to generate a full 13-frame photoreal walkthrough set for Carolverse Map, demonstrating it can hold one consistent style (warm parchment, gold, ink) across a multi-scene sequence rather than only one-off images.

New Capability2026-07-29

Image Generator now applies an agent's registered look from the registry when generating images.

Fix2026-07-27

2025-04-01: Rotated the Gemini image-generation key to a funded project, fixing the 429 error. The service was restarted and generation now works.

Fix2026-07-26

The image generator now receives the depicted agent per scene so it can inject the correct canonical reference face. This corrects previous behavior where it always used Scriber's reference face.

🗂️Tabs & Screens

Tab inventory is being built — see CAROL-INI-077 step 7.

👤Owner

Scriber · Chronicler of the Build Pipeline

📚Recent initiatives

Initiatives that touched this app — a short summary each; open one for the full story.

CAROL-INI-3757-00: Image generator: a payer-less refusal must say the caller was DEFAULTED to Scriber's track
2026-08-10 incident: an operator hand-call for a Logbook hero named no agent_id/bill_to, was defaulted to Scriber's initiatives-creative-media per ruling 1226 (CAROL-INI-3736), an\u2026
Orion · 2026-08-12 18:53
CAROL-INI-3339-00: Rotate the Gemini image-generation key to the funded project and restart the image generator
The Gemini image key was on a depleted prepay project (429). Ninad supplied a funded key. Rotate GEMINI_API_KEY in the env and restart the image generator so it reloads; verify ge\u2026
Orion · 2026-07-27 18:49
CAROL-INI-1818-00: Logbook blogs feature the right agent faces — photoreal reference sets + character bibles for key agents, generalize the image generator beyond Carol/Orion, and give the Author agent roles + per-image agent selection
Today the image generator can only inject Carol or Orion photoreal faces; every other agent falls back to Orion. So blog images cannot show the agents a post is actually about. Bu\u2026
Orion · 2026-06-18 16:36
Browse all initiatives →