Resolution is the single largest lever on an AI video bill. Not the model, not the provider β the number of pixels you asked for.
At one provider, the same model costs five times more at 1080p than at 480p. And because providers tier resolution differently β some charge per tier, some charge one flat rate β the cheapest provider for your job changes depending on which tier you pick.
Prices last checked: August 26, 2026. Live figures are on the compare page.
The cost curve inside one model
Seedance 1 Pro at Atlas Cloud, priced per second of output video:
| Tier | Per second | 5-second clip | 200 clips |
|---|---|---|---|
| 480p | $0.022 | $0.11 | $22 |
| 720p | $0.047 | $0.235 | $47 |
| 1080p | $0.110 | $0.55 | $110 |
480p to 720p roughly doubles the price. 720p to 1080p roughly doubles it again. Five times from bottom to top, for the same model, the same prompt and the same clip length.
The pattern repeats across families. Kling 3.0 at Atlas Cloud:
| Tier | Per second | 5-second clip |
|---|---|---|
| Standard | $0.071 | $0.355 |
| Pro | $0.095 | $0.475 |
| 4K | $0.357 | $1.785 |
Standard to Pro is a 34% step. Pro to 4K is another 276%. The 4K tier is not an incremental upgrade; it is a different budget.
And on fal.ai, MiniMax H3 publishes four tiers on one page: "$0.05 per second at 480p, $0.06 per second at 768p, $0.13 per second at 2K and $0.16 per second at 4K."
Where it gets interesting: providers tier differently
Some providers charge a different rate per resolution. Others charge one flat rate whatever you ask for.
Seedance 1 Pro, same model, two providers:
| Resolution | Atlas Cloud | Replicate | Cheaper |
|---|---|---|---|
| 480p | $0.022 | $0.030 | Atlas Cloud |
| 720p | $0.047 | $0.030 | Replicate |
| 1080p | $0.110 | $0.030 | Replicate, by 73% |
Replicate publishes a single rate β $0.03 per second of output video β with no resolution tier at all.
So the answer to "which provider is cheaper for Seedance 1 Pro" is: it depends entirely on your resolution, and the answer flips between 480p and 720p. At 480p Atlas Cloud is 27% cheaper. At 1080p Replicate is nearly four times cheaper.
If you picked a provider by testing at 480p and then shipped at 1080p, you chose wrong and the test told you nothing.
The tiers you will meet
| Provider | Model | How resolution is priced |
|---|---|---|
| Atlas Cloud | Seedance 1 Pro | separate endpoint per tier: 480p, 720p, 1080p |
| Atlas Cloud | Kling 3.0 | separate endpoint per tier: Std, Pro, 4K |
| Replicate | Seedance 1 Pro | one flat rate, no tier published |
| fal.ai | Wan 2.7 | $0.10 at 720p, $0.15 at 1080p |
| fal.ai | MiniMax H3 | $0.05 / $0.06 / $0.13 / $0.16 across four tiers |
| fal.ai | Happy Horse 1.1 | $0.14 at 720p, $0.18 at 1080p |
One caution on the "no tier published" rows. It can mean the provider genuinely charges one rate regardless β or that they have not documented a difference. Those are different facts, and only the first is safe to build a budget on. When we cannot tell which, we say so rather than assume.
The two multipliers people forget
Clip length. Every price above is per second. Ten-second clips cost exactly twice what five-second clips cost. A brief change from "5 seconds" to "10 seconds" doubles a whole project.
Takes. Video prompts rarely land first time. If you generate three takes per finished clip, every number in this article triples. Measure your real reject rate during a pilot rather than assuming it away.
Together these two multiply. Ten-second clips at three takes each is six times the naive estimate β before you have changed model or resolution.
Which tier do you actually need?
- Social and mobile-first. 480p or 720p. Most feeds downscale hard, and at 480p on Seedance you are paying $0.11 for a five-second clip.
- Web hero video and product pages. 720p is the sweet spot. It is where most providers' pricing is still reasonable and where the visible quality gap to 1080p is smallest.
- Anything that goes full-screen on a desktop monitor. 1080p, and budget for it β roughly five times the 480p line.
- 4K. Only when the delivery spec demands it. At $0.357 per second, a single five-second clip costs $1.79; two hundred of them is $357. Consider generating at 1080p and upscaling β Atlas Cloud lists a video upscaler at $0.018 per second, which is a fraction of the 4K generation premium.
And pick your provider after your resolution, not before. The Seedance table above is the whole argument: the same model, the same clip, and the cheaper provider flips depending on a decision most teams make late.
Explore more on ElliSekiz
Compare video prices in the same unit. Every model and provider, each price in the unit that provider bills in, on the compare page.
Test at the resolution you will actually ship. The video models index lists the catalogue, and every model page has a playground β generate one clip at your real target before committing to two hundred.
Sources
Read from the providers' own sources on August 26, 2026.
- Atlas Cloud model catalogue API,
https://api.atlascloud.ai/api/v1/models - Replicate β Seedance 1 Pro
- fal.ai β Wan 2.7 Text-to-video
- fal.ai β MiniMax H3 Text-to-video
- fal.ai β Happy Horse 1.1 Text-to-video
