AI video model pricing — real numbers
Most model comparisons quote vendor list price and call it a day. We list the credits and USD cost we actually charge for each generation, by model and resolution.
Pricing
Cost per generation
30-day rolling actuals once we accumulate enough generations. Until then: list-price reference, with "updating with real production data" labels.
| Model · pricingKey | Runs (30d) | Success | Avg USD | Avg latency |
|---|---|---|---|---|
| grok-imagine/image-to-video | 51 | 78% | $0.300 | 28.6s |
| grok-imagine/text-to-video | 2 | 100% | $0.300 | 44.8s |
How we calculate
Methodology
Source
We read aggregated rows from generation_analytics — the table populated on every successful and failed job. No raw prompts; PII-redacted.
Cost basis
kie.ai-published per-generation rate × duration. Refunded jobs subtract from the daily total.
Refresh cadence
Weekly, after deploys. The table is server-rendered.
Pick the right model in the generator
Cost preview shows up before you spend.