AI Video API Pricing Compared (October 2026): Per Second vs Per Video

List prices, October 2026

Prices below are public list rates from each provider or its published pricing page at the time of writing. They change often; follow the sources at the end before you commit.

Model	Provider	Price	10-second clip
Kling 3 Standard	fal	$0.084/s (no audio), $0.126/s (audio)	$0.84-$1.26
Kling 3 Pro	fal	$0.112/s (no audio), $0.168/s (audio)	$1.12-$1.68
Kling 3 Standard / Pro	Replicate	$0.071/s / $0.095/s	$0.71-$0.95
Veo 3.1 Lite	Google Gemini API	$0.05/s (720p), $0.08/s (1080p)	$0.64 (8s, 1080p)
Veo 3.1 Fast	fal	$0.10/s, $0.15/s with audio	$1.20 (8s, audio)
Veo 3.1 Standard	fal	$0.20/s, $0.40/s with audio	$3.20 (8s, audio)
Gemini Omni Flash	Google Gemini API	~$0.10/s at 720p (token-billed)	~$1.00
Seedance 2.0	BytePlus	$0.07/s (480p) to $0.37/s (1080p), $0.78/s (4K)	$0.70 (480p) to $3.70 (1080p)
Seedance 2.0 Fast / Mini	BytePlus	from $0.06/s / from $0.04/s (480p)	from $0.40
MiniMax H3	MiniMax, OpenRouter	$0.13/s	$1.30
Wan 3.0	Alibaba Model Studio	$0.05/s (480p) to $0.20/s (1080p)	$0.50-$2.00
Wan 2.5	fal / Replicate	$0.05/s / $0.071/s	$0.50-$0.71
Grok Imagine Video 1.5	xAI	$4.20/min (~$0.07/s)	~$0.70
Runway Gen-4.5 / Gen-4 Turbo	Runway API	$0.12/s / $0.05/s	$1.20 / $0.50

The three billing models

1. Per second of output

fal, Replicate, Runway, xAI and MiniMax bill each second of generated video, often with surcharges for audio, higher resolution or "Pro" tiers. It is transparent and has no commitment, but costs scale linearly: double the length, double the bill. Runway also requires a $10 minimum top-up and bills API credits separately from app subscriptions.

2. Tokens

Google's Gemini API (Gemini Omni) and BytePlus (Seedance) bill tokens, which are then converted from resolution and length. Fine-grained, but harder to forecast: you have to estimate tokens per second before you know the cost of a clip.

3. Per video, inside a plan

VideoGenAPI includes most models in a monthly plan and counts videos, not seconds. A 10-second clip uses the same allowance as a 5-second one. Two premium Google models are billed per video on top: Veo 3 Fast at $0.45 and Veo 3.1 Fast at $1.50.

What 1,000 ten-second clips cost per month

Setup	Monthly cost
Kling 3 Standard on fal, no audio ($0.084/s)	$840
Kling 3 Standard on Replicate ($0.071/s)	$710
MiniMax H3 at list ($0.13/s)	$1,300
Seedance 2.0 at 1080p on BytePlus ($0.37/s)	$3,700
Kling 3, MiniMax H3 or Seedance 2.0 on VideoGenAPI Starter (4,500 videos included)	$49

Plan limits apply: monthly video allowance, resolution available on each model, and rate limits. Run your own mix in the pricing calculator.

When per-second billing is the better deal

Very low volume. Twenty 10-second Kling clips on fal cost about $17, less than the $29 entry plan.
You need a provider-only feature, such as 4K Seedance 2.0, Veo 3.1 Standard or brand-new models before aggregators add them.
Spiky, unpredictable usage where a monthly commitment would sit idle.

When per-video billing wins

Steady volume above a few dozen clips a month.
Longer clips, where per-second billing hurts most.
Several models: one plan instead of separate balances at Google, BytePlus, MiniMax and Kuaishou.
Predictable budgets for clients or finance, without converting tokens to seconds.

For a provider-by-provider view beyond price, read Best AI video generation APIs in 2026, and see the fal.ai and Replicate comparisons.

Frequently asked questions

How much does an AI video API cost per second?

Between about $0.03 and $0.78 per second of output in October 2026, depending on the model, resolution, audio and provider. Most popular models sit between $0.07 and $0.20 per second.

What is the cheapest AI video generation API?

At list price, entry tiers such as Seedance 2.0 Mini, Veo 3.1 Lite at 720p or Wan models start around $0.03 to $0.05 per second. At steady volume, a flat plan billed per video, such as VideoGenAPI at about $0.011 per video on Starter, is usually cheaper.

How much does a 10-second Kling 3 video cost?

About $0.71 to $1.68 on per-second providers depending on tier and audio. On VideoGenAPI it counts as one video in your plan.

Why do some providers bill tokens instead of seconds?

Google and BytePlus reuse their language-model token billing for video. Tokens depend on resolution and length, so you have to convert them to estimate the price of a clip.

Sources

Third-party facts and prices were checked on the date of publication and can change; check the provider before buying.