Every page in this catalog describes a format. This page describes the machinery underneath all of them, because it is the same machinery each time and it is the part that decides whether the output is usable.
What happens after you name a format
You do not write a prompt for this. Hermoso reads your brand first — the product, the logo, the palette, the voice — from your website or a short description, then plans the ad against this format: a beat list, the on-screen copy, and the casting if a person appears. The render happens beat by beat, and every word that ends up on screen is composited afterwards by the app rather than painted by the model, so headlines, captions and the end card come out exactly as written instead of warped and half-invented.
If you have a real product photograph, add it. It is the difference between an ad showing your packaging and an ad showing packaging a model imagined, and it is the single biggest quality lever on any product format here.
Why the text is composited rather than generated
Image and video models draw letters. They do not spell them. Ask one for a headline and you get type that is nearly right: a doubled letter, a word that does not exist, kerning that drifts halfway through. On a product package the same failure invents a variant you do not sell.
So no format in this catalog asks a model to write. The picture is rendered first, and the headline, the supporting line, the captions, the step numbers, the prices and the end card are laid on afterwards by the app as real typeset text. That is why a bundle ad can carry figures that add up and a step-by-step demo can carry step numbers in the right order. It is also why brand colours are specified by name rather than by hex code anywhere in the pipeline: a model handed a hex code will sometimes paint the code onto the picture.
The one input that changes the output most
A real product photograph. With one, the render is anchored to your actual packaging, your actual label and your actual colourway. Without one, the model invents packaging that looks like your category rather than like your product, and on a premium item invented packaging reads as counterfeit.
Several formats will not run at all without assets, and say so before they start rather than failing part-way through: the AirDrop carousel needs three product photos, the moodboard grid needs four, the app tour needs at least two real screenshots of your interface.
The two rendering paths
Most of the catalog is model-rendered: a planner writes the ad against the format and a video or image model produces the frames. The native ad templates are the other path, and nothing in them is generated at all. Each is a self-contained HTML page stepped frame by frame in a headless browser and assembled with ffmpeg, so the same inputs produce the same file and the copy is real browser-rendered text. That path costs a couple of credits rather than a video generation, which is what makes it practical to test ten wordings of the same line.
Driving a format from an agent
Every format on this site is reachable from an AI agent as well as the web studio. Hermoso ships an MCP server, so Claude, Claude Code, Cursor, Codex and any other MCP client can call plan_ad to draft the creative and render_ad to produce it, name a format as the recipe, and get a finished file back — with the same brand grounding and the same composited-text guarantees. The MCP and CLI overview covers the setup, and the generated tool reference is the complete list.
Where to go next
The creative recipes index lists all 141 formats by family. Native ad templates covers the 13 deterministic HTML formats, and explainer styles covers the 18 looks available to a narrated explainer film.
Point Hermoso at your brand and render one of these formats today — the free plan needs no card.
Start free → See pricing