Skip to main content
Select a Brand Kit in Studio chat, or @mention one, to give the agent its identity and guidelines. Brand selection is saved for that conversation. For an independent film or other work without brand guidance, set Brand kit → None and attach the individual references you need. Workspace rules and lightweight brand summaries can still be available; None does not create an isolated project library.

​
Browsing reference assets

In Assets, the Characters, Images and Videos sections share the Collections list layout: a thumbnail, name, supporting text and updated date, with the existing actions beside the row. Choose List / Grid next to search; that preference is shared with the Brand Kits overview. Image and character rows open their editor. Uploaded video rows retain their video preview and action menu.

​
@mentioning a Brand Kit

In the Studio chat, type @ followed by your brand kit name. The agent loads your brand context:
“@Acme Brand — create a hero image for our spring campaign”
Mentions appear as styled chips with thumbnails in your message. The agent confirms what it loaded before generating.

​
What the agent sees

When you @mention a brand kit, the agent gets:
  • A catalog of linked logos and reference images, with a limited selection shown as visual context
  • Your defined color palette
  • Your typography preferences
  • Your full Voice & Rules document (voice, personality, banned language, audience cues, hard rules)
  • Any skills associated with the kit, if invoked in the same turn
It uses all of this as creative constraints — not decoration. “Always use natural lighting” in your Voice & Rules means the agent won’t generate studio-lit shots unless you override.

​
Scene references in briefs

In a Creative Brief, choose a scene and type @ to attach an available asset, or use the + below the prompt. Its inline chip and thumbnail represent the same scene reference. Removing the thumbnail removes its mentions from that scene; deleting the final inline mention also detaches it. Other scenes keep their selected assets, and the original remains in your library. Reference labels are locked inside prompts. Rename the underlying asset from its own editing surface. Color values such as #533AFD display a swatch while keeping the exact value in the generation prompt. See Creative Briefs for editing, output settings and saved-brief recovery.

​
Combining references with brand assets

Drag a mood board image onto the canvas and @mention your brand — the agent matches the reference’s style while staying within your brand’s visual language. Analysis data from the reference (lighting, composition, color) combines with brand constraints (palette, typography, guidelines) to produce work that’s both inspired and on-brand.

​
How photography is selected

The agent receives a catalog of kit photography and selects references for the request. Large catalogs can use photography search. Organizing a kit into collections does not currently restrict this catalog to one campaign’s collection.
  • Subjects and treatments — when you save a photo (or auto-extract from a brand book), Memo tags it with subjects (“hands”, “product”, “skyline”) and a treatment (“lifestyle”, “product”, “detail”, “editorial”, “portrait”). These power the search.
  • Semantic + keyword search — retrieval blends semantic embedding similarity with keyword matching, so “warm hands holding a mug” finds the close-crop hands shots even if the saved title says nothing about warmth.
  • Reference verification — before generation, the agent verifies the retrieved photos are actually relevant to your prompt. Off-target results are dropped instead of being forced into the output.
This is why a well-tagged photo library produces dramatically better generations than a dumping-ground one. Treat tagging as part of the kit setup, not an afterthought.

​
Color palette enforcement

Named brand colors should retain their exact values in brand design elements. They do not require recoloring skin, scenery, lighting, or wardrobe. State any alternative art direction explicitly. An invoked Skill’s explicit palette takes precedence over the kit palette. Using a saved character or product does not by itself request an added logo. Describe where branding should appear when you want it visible.

​
Saving generations as brand assets

From an image’s canvas context menu, Save to Brand Assets saves a reusable asset in the workspace. Saving it does not automatically choose a Brand Kit or kit collection. Add to collection opens the shared navigator for General collections and Brand Kits. Choose a destination, then check its collections. Kit collections accept product and style images, characters, graphic elements, photography, and uploaded videos; logos belong in the kit Overview. A collection’s name alone does not connect it to a Studio project or activate its contents as generation context. The system compounds. Generate → save → reference → generate again. Each cycle builds on the last.

​
Durable Character sheets

From a project’s assets page, move a file into Characters and choose which Character it belongs to. Memo saves a permanent copy with its own sheet identity and source record. The connected agent can do the same with promote_character_sheet after reading the file’s current Character-link revision. This permanent sheet survives removal of the source generation. When editing a Character, a stale sheet list cannot silently remove a newly added sheet. Refresh after a conflict, then make your intended edit. Removing a sheet from the refreshed list detaches it from that Character; its source record and permanent copy are retained for recovery. Automatic sheet labels only fill missing descriptions and preserve labels you supplied.