Price each image or video job before it starts

How to do it

One video request can cost more than thousands of text calls, and a retry loop multiplies it. Compute the price from the provider's rates for resolution, seconds and number of outputs, in the gateway's price table. Show the user the cost in credits before a large job. Charge the estimate against the daily budget in price every model call and cap the spend before the call, then settle the actual cost after. Run long video jobs as background jobs with a status, not inside a web request.

Examples

TypeScript example

export function estimateVideoUsd(job: VideoJob, rates: VideoRates): number {
  return job.seconds * job.count * rates.usdPerSecond[job.resolution];
}

if (estimateVideoUsd(job, rates) > MAX_JOB_USD) throw new JobTooExpensive();
await budgets.reserve(user.id, estimateVideoUsd(job, rates));

Python example

def estimate_video_usd(job: VideoJob, rates: VideoRates) -> float:
    return job.seconds * job.count * rates.usd_per_second[job.resolution]

if estimate_video_usd(job, rates) > MAX_JOB_USD:
    raise JobTooExpensive()
budgets.reserve(user.id, estimate_video_usd(job, rates))

How to check

A test submits a job priced above the per-job limit and one that would pass the user's daily budget. Both are refused before any provider call. A job under both limits records its estimated and its actual cost.

Sources

• OWASP LLM06:2026 Unbounded Consumption

Related rules

• AIF-42 Price every model call and cap the spend in AI features