How resolution changes what a video API costs

Updated 2026-10-02

Ask for the price of a video API and you will usually get one per-second number. That number belongs to a specific resolution tier, often the lowest one. Resolution is one of the two main drivers of cost (the other is duration), and the way it works is less uniform than most rate cards suggest.

Why resolution costs more

Higher resolution means more pixels to generate per frame, and typically more compute and more time per second of video. Providers pass that through as a higher per-second rate for higher tiers. The details of how they price are theirs, and we cannot see their cost structure. What you can observe is the pattern in the published rates, and that is what matters for planning.

The formula is simple:

job_cost = requested_seconds * rate(model, host, tier) * (1 + platform_fee)

On VideoRouter the fee is a flat 2% for image and video, the cost is charged once at creation from the requested duration, and polling is free. The pricing explainer covers that billing flow. This page focuses on the tier argument.

The live numbers

The table shows, for each model, the tier that is being compared, the cheapest and priciest host, and the percentage gap. Read the tier column first.

Per second, before VideoRouter's 2% platform fee. For tiered models each row compares the resolution tier with the widest host-to-host gap. Built 2026-10-02 from the live catalog.

Per-second, per-tier mechanics

• Each tier has its own rate. 480p, 720p, 1080p, 1440p and 2160p are different line items on the same model page.
• Tiers are not uniformly offered. A host may list only some tiers. A dash in the grid means it does not offer that tier, which also means it is not a failover target for that tier.
• Labels can hide variants. Tier names such as 1080p-esr denote enhanced-super-resolution variants that a host prices separately from the plain tier of the same number.
• Not every model is per-second. The docs call out models that bill a flat amount per clip or per request regardless of requested duration. For those, the tier-by-seconds model does not apply.
• Some options change the price too. Whether a model renders audio, or which input mode you use, can move the rate on certain hosts. The model page states the basis.

Why host rankings change between tiers

If you choose a host from the lowest-tier column and then ship a higher tier, you may be on the wrong host. Three reasons the ranking reorders:

1. Hosts price each tier independently, so a discount at one tier does not carry over.
2. Coverage varies. At a tier offered by few hosts, the range is narrow and the cheapest host may be a different one.
3. Rates do not scale by a fixed multiple from tier to tier.

So always read the column for the tier you will ship. The pricing page and each model's own page show the whole grid, and unpinned requests choose the cheapest healthy host for the request automatically.

Comparing across tiers fairly

To compare a model at two tiers, hold everything else fixed: same model, same host if you can, same duration. Then compute cost per clip at each tier and divide by the value you get. A practical way to think about it:

What this means for a buyer

Three habits follow from all of this. First, always quote a price with its tier attached, for example in your internal docs and vendor comparisons, because a bare per-second number is ambiguous. Second, include the tier in the cache key and the cost log of your application, so spend reports can be sliced by tier. Third, treat a tier change as a pricing event: it may also move you to a different cheapest host, so re-run the comparison instead of editing one constant.

A tier-aware budget

If part of your volume is draft and part is final, budget them as separate lines with their own tier and rate, instead of averaging:

total = (draft_clips * draft_seconds * rate_draft_tier * attempts_draft
       + final_clips * final_seconds * rate_final_tier * attempts_final) * (1 + fee)

The retry multipliers differ by stage, and drafts at a low tier have a much lower cost per attempt, which is why exploration is cheap there. The budget estimation guide walks through the full method.

Checking the tier you got

Because a mismatch is silent, add a cheap assertion to your pipeline: after each job completes, read the video's dimensions and log them next to the requested tier. If they differ, you will see it in your logs before you see it on your invoice. See the quickstart for the request fields, and live clip costs on the cost page. A key takes a minute at videorouter.sh/signup.

Frequently asked questions

Keep reading

• How AI Video API Pricing Works — Per-Second, Per-Tier, Per-Host
• Cheapest Video Generation API: How to Find It for Your Workload
• How to Estimate Your AI Video API Budget, Step by Step
• AI Video Generation Cost Per Minute: How to Calculate It