AI avatar video API pricing: cost per second and per minute

Sume prices AI avatar video per second, by quality tier and by whether the request sends a product image: $0.184 to $0.58 per second, plus a 5.5% agent fee by default. Per minute, meaning 60 × the per-second rate, that is $11.04 to $34.80 before the fee, and $14.70 on the default plus tier without a product image.

The per-second rates are read from the code that renders API pricing, the rate card the Developer API docs call the source of truth for prices, and each per-minute figure is 60 times the per-second rate. Request rules come from Generate avatar video, read on 2026-09-26.

How much does an AI avatar video cost per second and per minute?

Avatar video has three quality tiers, and each tier has two rates: one for a productless video and one for a video that sends product_image. quality accepts standard, plus, or max; omit it and you get plus.

For example, 30 × the default plus rate without a product image is $7.35. Which tier fits which job is covered in Talking avatar video API.

One request takes a script or video_inputs plan that Sume estimates at 4–60 seconds inclusive; AI avatar video longer than 60 seconds shows how to split a longer script into several jobs.

Is there a fee on top of the per-second rate?

Yes. Every rate in the table is before the agent fee: avatar video is billed at its per-second rate plus a 5.5% agent fee by default.

You do not need to do the arithmetic by hand. Every accepted talking-video submit returns a usage block with Sume's USD estimate for that request, described in the API reference:

• billable_amount_usd: dollars, the figure to quote. It is always billable_amount_usd_micros / 1_000_000.
• model_amount_usd_micros and agent_fee_amount_usd_micros, when present: the model amount, and the agent fee charged on top of it.
• billable_amount_usd_cents: ceil(micros / 10_000) cents for wallet rounding. Not dollars.

What else changes what I pay?

Beyond tier, length, and the product image, three choices move the total:

• Creating the avatar: $0.95 per avatar, one time. Every later video reuses its handle.
• Changing tier after a preview. generate-video on an avatar video preview can override quality for the final render, and the reservation uses that tier; Avatar video previews covers the flow.
• Inline captions. captions on the request is a separate add-on on the avatar-video estimate, not a separate billed caption job.

Do failed or canceled avatar jobs cost money?

A failed job is not charged, and neither is a job you cancel before generation starts: Sume reserves the estimated amount when it accepts the request, captures it when the job completes, and releases or refunds the reservation when the job fails. Once generation has started, a cancel returns 409 job_generation_already_started and the job completes or fails normally, per Generation admission. Do failed AI video generations cost credits? covers the whole lifecycle.