Working with your asset library¶

Your organization's asset library holds the characters, locations, looks, voices, props and products you have built. Anything published to the library can be reused in any production — you build it once, and every later production draws on the same identity, so a character looks and sounds the same everywhere it appears.

Building an asset¶

Each kind of asset has its own creation app in the studio. Pick the app that matches what you are making, and you will move through the same five steps:

Name it. This creates a draft. Nothing is generated yet, so nothing is charged.
Describe it. Write the description yourself, or use AI Assist to turn the name into a description worth generating from. You can rename a draft or rewrite its description at any point while it is still a draft.
Generate options. The studio produces several candidates. Pick the one you want — choosing between options you have already generated does not cost anything extra.
Record the rights. Say what you are cleared to use. An asset with uncleared rights cannot be published.
Publish. The asset joins your library and becomes available to every production.

You will see a cost quote before any step that generates something, so there are no surprises.

Using your own images¶

You do not have to describe everything from scratch. If the character is a real person, the location is a place that already exists, or the look was set somewhere else, upload the images — and be deliberate about how you attach them, because the two options behave very differently:

Reference — a mood board. It influences the result: palette, mood, silhouette, era. The generator is free to depart from it.
Canon — this is the asset. Not "something like this" — this. The generator starts from your actual pixels rather than from a description. Use canon for a real person or a real place.

Getting this wrong is the most common disappointment: uploading forty photos of one face as reference produces a face that is only ever nearby. If you need that exact face, mark it canon.

Attaching more images does not dilute them. Each input has a weight, and the strongest are the ones the generator actually receives — so a well-marked canon image outranks a pile of loose references rather than being lost among them.

Reading a published asset¶

Click any published asset to open its detail view: the reference image at a size you can judge, its rights, description, reuse count, and a Carries block that states plainly what the asset does and does not contain.

If an asset says "recorded, unreachable", its picture exists in the record but cannot currently be loaded. That is a different problem from an asset that never had one, and it is worth investigating rather than ignoring — the list marks these with ! so you can spot them without opening each asset.

Worlds¶

A world bundles everything belonging to one setting — its cast, its standing locations, its look, its voices — under a single name. When you start a new production you can instantiate it from a world, and it begins already populated. The studio tells you exactly what was brought in, so you can see immediately if a world contributed less than you expected.

Removing things¶

Published assets are archived rather than deleted. They disappear from your library and from the pickers, but productions that already used them keep their history intact.