VisionAI branding

Brand name and taglines for this customer. The brand name is rendered EXACTLY on visible labels, woven tags, and reflective-tape branding text in every generation. Taglines are passed to the analysis pass so ambiguous brand text is read with the intended spelling.

Studio lighting reference

One reference image per customer that defines their house studio lighting. Claude describes it as if briefing a gaffer; the description text is fed to Nano Banana when the customer generates against the studio default backdrop. When a custom environment is selected the environment image takes over and this reference is ignored.

Character profiles

System profiles (Peter, Albert) are shared across all customers. Customer-scoped profiles below show with 'CUSTOMER' in the Mode column. Click any customer-scoped row to see the reference images Claude used to build it. Add new ones with the button under the table.

Name	Mode	Created	Updated
Loading…

Expression gallery

Up to 20 facial-expression references. Claude describes each one as a casting director would — focusing on expression, head tilt, and gaze direction (not identity, hair, or clothing). The text replaces the default expression directive in the master prompt. Empty gallery = Nano Banana picks freely from the identity references. Same anti-repetition rule as poses.

Pose gallery

Up to 20 reference images that define this customer's pose vocabulary. Claude describes each one as a movement coach would, and that description is fed to Nano Banana on every generation. The reference image is NOT sent to the model — only the text — to avoid leaking garments or backgrounds. When the gallery is empty the built-in pose library is used instead. Anti-repetition: the most recent pose is excluded on the next generation so two runs in a row never share a pose.