Should Teams Choose per-Second or per-Request Pricing for AI Video APIs?

Overview

Compare the two units on an identical deliverable. As a dated per-second example, WaveSpeedAI’s google/veo3 page listed $0.20/s without audio on September 29, 2026, so an 8-second clip illustrates $1.60. Against a hypothetical $2.00 per-request offer for the same 8-second specification, that listed per-second example is $0.40 lower before rerenders. The $2.00 figure is a worksheet assumption, not another provider’s quote.

Define the same deliverable for both candidates: duration, aspect ratio, resolution, audio requirement and expected number of final variants. Record each provider’s current route-specific billing rules, then price short and long jobs separately. WaveSpeedAI’s public table shows per-second billing for selected video routes, but the price guide says model and inputs determine the estimate and final charge may differ. Do not extend one route’s unit to the entire catalog. For a provider quoting per request, check what a request buys and whether failed or partially completed tasks are charged under its terms.

Next, run a representative sample and calculate cost per approved second or approved clip. If a shorter render needs twice as many attempts to pass review, its unit price may not be the cheaper production choice. Include human editing and storage only when comparing full workflow economics, and show them as separate rows.

Pick the billing model that is predictable for your workload and contract. Recheck the live quote before changing customer prices; a comparison built on yesterday’s configuration may no longer hold.

Document the exact quote time and configuration used in the decision, then set a review trigger for material route, price or usage-pattern changes.

Source: WaveSpeedAI pricing; WaveSpeedAI Veo 3 route; WaveSpeedAI How Pricing Works. Source check: 2026-09-29.