No markup on models: Seedance, Kling, Veo and Nano Banana at the model maker's price
Hermoso now runs every image, video and voice model at what the model maker charges, with no markup. What changed, why we did it, and what it means when you make ads.
Starting today, Hermoso charges what the model maker charges for every image, video and voice model. No markup. Seedance, Kling, Veo, Nano Banana, GPT Image, ElevenLabs and every other model in the catalog run at the model price.
That includes the models working inside bigger jobs. When a finished video ad renders its clips, or when you upscale, reframe, dub or swap a voice, the model part of that job is priced at what the model maker charges too.
What changed
Like most AI creative tools, we used to put a markup on top of the model. Ours came out at roughly double the model price. That's a normal number in this industry, and it's how a lot of tools pay for everything else they do.
Now the model price is the price. Take Veo 3.1: Google charges $0.40 for each second of video, so an 8-second clip with sound costs $3.20 to run. Seedance 2.5 at 720p runs about $0.23 a second. Nano Banana 2.1 makes an image for about 4 cents. Those are the numbers your price is built from now, and the app shows you the exact price before every render. To see what each of them makes for the money, we ran the same ad on ten video models: 10 AI video models, one prompt.
Our prices follow the model makers' prices. When one of them cuts a price, you get the cut too.
Why we did it
A markup is a tax on testing
Ads are won by testing. The brands that find a winner are usually the ones that tried more hooks, more angles and more takes than everybody else. Every one of those tries runs a model. When each try carries a markup, you quietly try less, and the ad you never made can't win.
We want to earn our keep on the work around the model
A model makes a clip. Hermoso does the rest of the job: it finds the ads already winning in your market, plans the ad, keeps it on brand, picks the right model for each shot, publishes to your channels and reads back how it did. That's the part we built, so that's the part we want to be paid for. Reselling a model at twice its price never felt like the product.
Anyone can look up the price
Model makers publish what they charge. You can find what a second of Veo costs in about a minute. A markup only works while nobody checks, and that's a strange thing to build trust on when people hand you their marketing.
What it means when you make ads
- More takes for the same budget. Render three versions of the opening shot and keep the best one, instead of settling for the first.
- The premium model when the shot needs it. Veo 3.1 for someone talking to camera, Seedance 2.5 at 1080p for the hero product shot. Picking the best model for the job gets a lot easier when it isn't marked up.
- Pennies where quality matters less. Hook tests on Veo 3.1 Fast and static variations on Nano Banana 2.1 cost cents each.
Here's a realistic week of testing for one product, at model prices:
| What | Model | Each | Total |
|---|---|---|---|
| Five 8-second hooks | Veo 3.1 Fast | $0.80 | $4.00 |
| Three takes of a 15-second product clip | Seedance 2.5, 720p | about $3.50 | about $10.50 |
| Twenty static ad variations | Nano Banana 2.1 | about 4 cents | about $0.85 |
| The week | about $15 |
That's a full creative test for about the price of lunch. The agent's planning, captions and editing add a little on top, but in ad creative the model is where the money goes, and that part now has no markup. For more budgets like this one, see what an AI video ad costs to make.
What stays the same
Every plan still gets the full studio: competitor ad research, image, video and UGC avatar ads, publishing and scheduling to your social channels, campaign management on your ad platforms, and unlimited profiles and team seats. Nothing is tier-locked. And Hermoso still works wherever you do: in the app, or from Claude, ChatGPT, Cursor and other AI agents through MCP.
Try it
Plans are on the pricing page. If you want the longer story of how AI tools price models, and how to compare two tools on equal terms, we wrote that up too: AI video generator pricing, explained.
Questions people ask
What does no markup on models mean?
The model part of every render is priced at what the model maker charges, not at a multiple of it. That covers every image, video and voice model on Hermoso, including the model runs inside finished ads, upscaling, reframing, dubbing and voice swap.
Which models have no markup?
All the image, video and voice models: Seedance, Kling, Veo, Gemini Omni, Hailuo, Nano Banana, GPT Image, FLUX, Seedream, ElevenLabs and the rest of the catalog.
Does the AI agent's own work have no markup too?
No markup on models covers the image, video and voice models. The agent's own work, like chat, planning and research summaries, and ad research data keep their usual prices.
What do the models cost?
As of October 2026, Veo 3.1 is $0.40 a second, Seedance 2.5 about $0.23 a second at 720p, Kling 3.0 about $0.14 a second and Nano Banana 2.1 about 4 cents an image. The app shows the exact price before every render.
What happens when a model maker changes its price?
Our prices follow the model makers' prices, so when a model gets cheaper, your price drops with it.