Base image

How to set the primary base image of a Series Lab series, and what Vision Assist suggests.

What it is for

The base image is the sample every NFT in the series resembles: composition, features, angle, style. The generator dresses variable traits onto it. In the interface the primary role is called the primary base image.

How to add one

Upload a PNG, JPG or WebP. A square from 1024×1024 works well. One clear subject, a quiet background, no text, logos or watermarks.
Or describe the character and generate a candidate: from scratch, or from the primary you already chose. The text is checked for brands before generation.
In the input/ gallery, set the best frame as the primary base image.

Vision Assist

The button on the Base tab reads the primary base image and suggests a style lock for the Constructor: angle, light, mood, framing, palette — plus starter trait hints (who the character is, the background, key attributes). With a Vision key (OpenAI or Claude) the read is more detailed. Without a key you still get a simple free analysis.

You can carry the suggestions into the Constructor and edit them. They do not replace your traits on their own.

Other files in input/

In the interface	File	Why
Primary base	`primary.png` (also jpg, webp)	The reference for the whole series
Line base	`base-<line>.*`	A separate reference for one character line
Style reference	`style-*`	Light or style alongside the primary
Extra	`extra-*`	A spare frame or a detail

Replace and delete

When you assign another primary, upload a file onto a name that is taken, or press Delete, the previous file goes to the trash. Anyone with access to Admin → Resources can restore it. The trash keeps files for 30 days by default.

The bottom layer

That same primary base image is the bottom layer for Compose layers on Generate. Files in layers/<trait>/<value>.png stack on top of it. Prompts stay. Composing does not charge credits.

One strong primary base image keeps a series more even than many unrelated references.