← Back to blog
AiAbout 6 min read

Forty-Four Story Cards and a Per-Second Bill: Higgsfield Turned Ads Into Inventory

Published Nov 8, 2026
Forty-Four Story Cards and a Per-Second Bill: Higgsfield Turned Ads Into Inventory

On 6 November, Higgsfield added Micro Drama Ads to Ads Studio, the static ad tool it opened in October. The feature does one thing: it turns a product into a three-beat story. A character has a problem, something goes wrong in public, a rival makes it worse, and the product arrives at the turning point.

The structure is a borrowed one: the short drama formula, compressed into a vertical ad, which has been the most reliable format in Chinese short-form video for three years. What is new is that the formula now ships as 44 named templates with a finished example ad attached to each one, and an agent that slots your product into whichever template you pick.

The catalog is the product

Eight styles at launch. Two are photoreal: Micro drama, described as live action shot like a film, and Cinematic story. Six are stylised: Cartoon story, Claymation story, Talking object, Inside the body, Doll story, Crochet story.

The story cards are not evenly distributed. Micro drama carries 18 of the 44, which tells you where the demand is. Cartoon story has 7, Claymation 4, and the rest hold 3 each. Each card is a template with a finished example, and a Recreate button reruns the plot with your product. Higgsfield says it checked the styles and story cards on 6 November.

A vertical strip of blank film frames in a steel holder, one position empty.

Read the card list and you can see the commercial logic. Public Failure is a hero who fails in public because of the problem the product solves. Bear Proof is an absurd problem narrated by its hero until the product ends it. Frizzilla makes the product a vain talking character who fixes a flaw twice so it lands on the rival. Inside Confession tells the story from inside the body, which is a supplement ad. Shrink the Creature turns the problem into a clay creature that grows until the product shrinks it.

Every one of those is a slot: problem shape, antagonist, resolution. Which means the creative decision moves earlier in the process. You are no longer writing a script, you are choosing which of 44 plots your product can survive being placed into, and then writing the twenty seconds around it.

The workflow, and the switch worth flipping first

Setup starts with a brand. In Tools, then Brands, paste your website URL and press Create my brand. Ads Studio fills the brand kit and product list from the site, or you build the kit manually if there is no site.

Then you describe the ad or paste a script, choose a style, a product from the kit, and an avatar. The avatar is the piece that holds a campaign together: one face across every cut, so ten versions look like ten ads from one brand rather than ten ads from ten brands. REF adds a reference image if the description is not enough.

Then there is the Allow menu, and this is where I would spend my first thirty seconds. The default is Generate without asking. You can switch it to Ask before generating. Everything downstream of that switch costs credits, and an agent that starts renders on its own initiative will start them at 3am.

Captions are not added by default. The cards ship vertical at 9:16, and if you call the same catalog from Claude or ChatGPT through the Higgsfield MCP, the /microdrama-ads command will take 16:9, 1:1, 4:3, 3:4 or 21:9 on request, at 480p, 720p or 1080p, with subtitles only if you ask. There is no stated cap on length.

The arithmetic

Higgsfield publishes per-second rates: 3 credits at 480p, 7 at 720p, 16 at 1080p. One credit is roughly $0.049, based on $49 for 1,000.

So a 60-second film at 1080p is 960 credits, about $47. Images, voiceover, music and agent requests are billed separately. In Ads Studio, each request to the agent also consumes credits unless the model is set to Auto, which is free. Four variations of a 30-second spot at 1080p is 1,920 credits before you have fixed anything, and the first version of a script is rarely the one that ships.

Put that against the rest of the market and the position is clear. On OpenRouter this week, video generation spans a 34x range: TBC's mischief 1 at $0.005 a second, HeyGen Video at $0.01, Wan 3.0 at $0.0425, MiniMax H3 Max at $0.05, Gemini Omni 1.2 Flash at roughly $0.07 for 720p, Seedance 2.5 at $0.1028, FLUX 3 Video at $0.17. Higgsfield's 1080p rate of about $0.78 a second is well above all of them, and it is not selling seconds.

What the markup buys

The model is not the product here. The brand kit, the 44 plots, the persistent avatar, and the agent that assembles a cut from a brief are the product. Buying seconds from an API gets you footage. Buying this gets you footage shaped like an ad, with a product placed at the turning point and a face that stays the same.

That is a real service and it has a real ceiling. The ceiling is continuity. One avatar solves the face. It does not solve the same rival returning in episode forty, the same kitchen across twenty spots, the same light in every exterior. Higgsfield's answer to a series is a repeated template, and a repeated template is exactly what audiences learn to skip.

Why this lands now

Two things changed in the last month. Story-shaped ad production moved from "an agency takes six weeks" to "a brief takes a conversation", and the tooling for that now arrives pre-populated with plots instead of a blank prompt box.

The second change is the MCP path. The same catalog runs inside Claude and ChatGPT, which means the person writing the brief and the tool producing the ad are in the same window. That is a workflow change disguised as an integration, and it is the reason a template library is more defensible than a model wrapper: the templates are content, and content accumulates.

It also changes who does the work. A marketer now writes the brief in a chat window instead of a creative director writing it in a treatment document. That is faster and it removes the step where somebody asks whether the idea is any good. The 44 cards are, in effect, the answer to that question, pre-answered by whoever built them. What is left is picking one and filling in the product.

The thing to check before you scale

Ask what happens to the credit bill when the first ad is bad. The question worth asking is what the correction path costs. A change of style is a new render. A change of script is a new render. A change of avatar invalidates the set, because the face was the only thing holding it together.

The teams that get value out of this will be the ones that write a brief detailed enough to survive the first render: a specific mishap, a specific location, a specific second where the product enters. The teams that burn budget will paste a product URL, press generate, and hope the agent fills the gap with something better than an office, a smile and sunlight.

Related articles