cast.Make a video · Free

How much does AI video cost per second in 2026?

As of 2026-10-18, listed prices for 720p AI video run from about $0.05 a second (Google's Veo 3.1 Lite on the Gemini API) to over $0.40 a second for premium models[1][5]. Most mainstream 720p options land between $0.08 and $0.30 a second, so a 10-second clip costs roughly $0.80 to $3, and 1080p often costs two to two and a half times as much as 720p on the same model[3].

Every major option, in dollars per second

Providers quote prices in different units: per second, per video, per million tokens or in app credits. The table converts each into US dollars per second of output. API prices are list prices without a video input. For subscription apps we divided the plan price by the seconds its credits buy, which assumes you use every credit; that is the best case, so real costs are higher if credits go unused.

USD per second of output as of 2026-10-18. "—" means the model or plan does not offer that resolution or the provider did not publish it.
Model and whereHow you pay720p1080pNotes
Veo 3.1, Gemini API[1]Per second$0.40$0.404K $0.60
Veo 3.1 Fast, Gemini API[1]Per second$0.10$0.124K $0.30
Veo 3.1 Lite, Gemini API[1]Per second$0.05$0.08No reference images
Seedance 2.5, BytePlus[3]Tokens$0.23$0.57From BytePlus's 5 s estimates[3]
Seedance 2.0, BytePlus[3]Tokens$0.15$0.374K offered
Seedance 2.0 Fast, BytePlus[3]Tokens$0.12—Fast tops out at 720p[4]
Seedance 2.0 Mini, BytePlus[3]Tokens$0.076—Cheapest Seedance
Seedance 2.5, fal.ai[5]Per second$0.473$1.164
Seedance 2.0 Fast, fal.ai[6]Per second$0.2419—
Seedance 2.5, Replicate[7]Per second$0.2312—$0.9676 with a video input
Seedance 2.0 Fast, Replicate[8]Per second$0.15—
Seedance 2.0 Fast, Atlas Cloud list[9]Per second$0.1548$0.3484After a 20% discount
Seedance on Atlas Cloud, our account's quotePer secondFast: $0.0582.5: $0.59Measured 2026-10-16, 9:16, audio on
MiniMax-H3, MiniMax API[10]Per second$0.08 (768P)$0.13 (2K)
Hailuo 2.3 legacy, MiniMax API[10]Per 6 s videoabout $0.047 (768P)about $0.082
Runway Gen-4.5, Standard plan[11]Creditsabout $0.29Resolution not stated
Runway Gen-4.5, Pro plan[11]Creditsabout $0.19Resolution not stated
Seedance 2.5 on Pika, Starter rate[13]Creditsabout $0.27—122 credits per 5 s
Wan 3.0 on Pika, Starter rate[13]Creditsabout $0.073—33 credits per 5 s
Luma Ray3.2, Plus plan[14]Creditsabout $0.06about $0.24
Veo 3.1 with audio on Luma Plus[14]Creditsabout $0.84280 credits per second
Seedance 2.0 Fast in Dreamina Basic[15]Creditsabout $0.079—Dreamina's own 19-video estimate[15]
Hedra Character 3 API (talking avatar)[17]Per secondfrom $0.025Animates a face over audio

Not in the table: Kling. Kling's membership and API price pages render only in a browser, so we could not verify Kling's own per-second prices for this update. OpenAI's Sora 2 API listed $0.10 a second before OpenAI shut it down on 2026-09-24[18]. Wan2.2's open weights are free under Apache 2.0[19], so you pay only for your own GPU time.

Why resolution and seconds drive the price

Video models do work in proportion to how many pixels they produce, for how many frames. BytePlus, ByteDance's cloud, states it outright: estimated tokens equal (input video seconds + output seconds) × width × height × frame rate ÷ 1024[3]. Take a 5-second 1280×720 clip at 24 frames a second: 1280 × 720 × 24 × 5 ÷ 1024 is 108,000 tokens. At Seedance 2.0 Fast's 5.6 USD per million tokens[3] that is about $0.60, which matches BytePlus's own estimate for a 5-second 720p Fast clip[3].

Everything else follows from that formula:

We go deeper on the economics, and on whether it will get cheaper, in why AI video is so expensive.

What we pay, and what Cast charges

We render through Atlas Cloud. Its free price endpoint quoted our account $0.058 a second for Seedance 2.0 Fast at 720p and $0.59 a second for Seedance 2.5 at 1080p on 2026-10-16. That is below Atlas's public list price for Fast, so treat it as one account's quote, not a price you can count on. The Seedance-only breakdown, with every provider's per-video price, is in how much Seedance costs.

Cast sells credits, and each video's credit price is set from what the render costs plus the prompt-writing model call and a margin. In dollars, a 5-second Standard video (Seedance 2.0 Fast, 720p) costs about $0.58 at the Starter pack rate and $0.47 at the Studio rate. A 10-second one is about $1.10 and a 15-second one about $1.62 at the Starter rate. A 5-second Pro video (Seedance 2.5, 1080p) is about $4.96 at the Premium rate. Your first 5-second video is free.

Turning any app's credits into dollars per second

  1. Find the plan's price and its monthly credits. Divide: that is your dollars per credit.
  2. Find the credit cost of one clip at the length and resolution you want, or the credits per second.
  3. Multiply, then divide by the clip's seconds.
  4. Ask what happens to unused credits. If they expire monthly, your real cost is higher than the table shows.

Worked example: Runway Standard is $15 for 625 credits, and Gen-4.5 uses 12 credits a second[11][12]. $15 ÷ 625 × 12 is about $0.29 a second. More on this in how AI video credits work.

What to pick when

Where Cast fits

Where it does not:

People also ask

What is the cheapest AI video per second?
Among listed API prices we found, Veo 3.1 Lite at $0.05 a second at 720p[1] and Seedance 2.0 Mini at $0.045 a second at 480p on Atlas Cloud[9]. Running open-weight Wan2.2 on your own GPU costs only the hardware[19].
Does 1080p cost more than 720p?
Usually two times or more, because 1080p has 2.25 times as many pixels: Seedance 2.5 on fal.ai is $0.4730 a second at 720p and $1.164 at 1080p[5]. Veo 3.1 is an exception, at $0.40 a second for both[1].
Am I charged for failed or blocked videos?
Usually not on APIs: BytePlus bills only successful generations[3], and Google says Veo charges apply only when a video is successfully generated[1]. Cast refunds failed renders automatically.
Does audio cost extra?
On some platforms. Google Cloud lists Veo 3.1 at $0.40 with audio and $0.20 video-only[2]. Our Atlas Cloud quotes for Seedance in the table were measured with audio on.
Why do apps charge more than the API price?
They add the cost of prompt-writing models, storage, payments, failed renders they refund, and a margin, and their credits often expire. Compare with how AI video credits work.

Sources

Claims about other companies come from their own pages, read on the date shown. Products and prices change; check the link for the current version.

  1. Google: Gemini API pricing (read )
  2. Google Cloud: generative AI pricing (read )
  3. BytePlus ModelArk: model pricing (read )
  4. BytePlus ModelArk: Seedance 2.0 (read )
  5. fal.ai: Seedance 2.5 text-to-video (read )
  6. fal.ai: Seedance 2.0 Fast text-to-video (read )
  7. Replicate: bytedance/seedance-2.5 (read )
  8. Replicate: bytedance/seedance-2.0-fast (read )
  9. Atlas Cloud: Seedance 2.0 prices (read )
  10. MiniMax: API pricing overview (read )
  11. Runway: pricing (read )
  12. Runway Help: how do credits work (read )
  13. Pika: pricing (read )
  14. Luma AI: pricing (read )
  15. Dreamina: Seedance 2.0 (read )
  16. Dreamina: home page (read )
  17. Hedra: models (read )
  18. OpenAI: Sora 2 model page (read )
  19. Wan2.2 on GitHub (read )

Related answers

Part of the Cast answers library (Prices and costs).

Updated