Annotated callouts

The product annotated with handwritten-style callout notes and arrows pointing at specific features. It reads like someone marked up a photo.

When to use it

Products where the details are the sell but not obvious in a photograph: materials, construction, hidden features. The handwriting makes it feel like guidance rather than advertising.

The thing to get right

Point at things that are visible. An arrow indicating a feature that cannot be seen in the frame draws attention to its absence.

The format at a glance

• Output: a still image ad.
• Structure: a single frame, with the copy composited on top.
• Family: Image ads.
• Recipe id: callout_notes, which is what you name when you ask for it by hand or from an agent.

What should the arrows point at?

Something visible in the frame. An arrow indicating a feature that cannot be seen draws attention to its absence, which is the opposite of what an annotation is for. Reframe the shot rather than adding a label.

Related formats

Static, direct-response creative for the feed: the headline, the offer, the proof, the comparison. Every word on these is composited by the app after the picture is rendered, so the copy comes out exactly as written.

• Native / organic post: an ad disguised as an organic native social post that blends into the feed with a low ad vibe
• Lifestyle + stat overlay: a lifestyle photo with bold stat or number overlays, proof composed onto a real moment
• Whiteboard explainer: a hand-drawn whiteboard or sketch explainer that sketches out the concept or how it works
• Bold color-block graphic: a bold geometric color-block design, the product popped on flat brand-colour blocks
• UGC side-by-side: a split-screen UGC comparison of two real moments side by side (problem vs solution, or before and after in use)
• Editorial / magazine: an editorial magazine cover or spread layout with premium, type-led fashion-mag energy

The full catalog is on the creative recipes index.

Models that render it

This format is rendered by an image model, with the copy composited on top afterwards. GPT Image 2.5 is the default pick. The image models page compares the rest, including Grok Imagine 2.0, MAI Image 2.5 Pro and Grok Imagine.

Hermoso plans and renders an annotated callouts from your brand rather than from a prompt you write, composites every on-screen word after the render so the copy comes out exact, and exposes the same format to an AI agent over MCP. How Hermoso renders a format covers all three.