November 1, 2026 · NascoreAI team
Ask five AI video tools what a clip costs and you will get five answers in five different units: per second, per credit, per generation, per subscription tier. This page converts the published prices into the one unit filmmakers actually buy — one 10-second clip, at a stated resolution, with audio on — and then shows the math that turns a raw generation cost into a retail price.
All figures below are published API list prices checked on 1 November 2026. They are generation costs — what the model charges the platform using it — not retail prices. The gap between the two is explained further down.
| Model — route | Published rate | Cost per 10s | Sustainable retail |
|---|---|---|---|
| MiniMax H3 — official API (768P) | $0.08/s | $0.80 | $1.15 |
| MiniMax H3 — self-hosted resellers (768P) | $0.03/s | $0.30 | $0.45 |
| PixVerse V6 — own API (720P, audio) | $0.053/s | $0.53 | $0.79 |
| Wan 3.0 — Alibaba Cloud (720P) | $0.10/s | $1.00 | $1.39 |
| Kling 3.0 Standard — via fal (720P, audio on) | $0.126/s | $1.26 | $1.79 |
| Kling 4.0 — via fal (720P) | $0.157/s | $1.57 | $2.19 |
| Seedance 2.5 — BytePlus official (720P) | $0.231/s | $2.31 | $3.19 (floor) |
| Runway Gen-4.5 — API (720P, no native audio) | $0.12/s | $1.20 | — |
| Veo 3.1 Standard (720P) | $0.40/s | $4.00 | — |
Two details move these numbers a lot. First, audio: Kling 3.0 Standard is $0.084 per second silent and $0.126 per second with sound — a 50% difference hiding inside the same model name. Second, resolution: the same model at 480P can cost half its 720P rate, which is why honest comparisons always state the resolution.
A platform cannot sell at generation cost, because not every generation is usable and payments are not free. The back-of-envelope we use adds a 25% retry uplift (roughly one re-roll in four) and divides by 0.93 to cover payment and platform fees of about 7%:
Sustainable retail ≈ generation cost × 1.25 ÷ 0.93
Run MiniMax H3 on the official API through it: $0.80 × 1.25 ÷ 0.93 ≈ $1.08 — which is why our Final video retails at $1.15, with a margin of a few cents. Run the same model through a self-hosted route at $0.03 per second and the floor drops to about $0.40, sustaining a $0.45 retail. Same model, same output — the route is the price.
At least one competitor advertises a flat $0.45 per video, described as a full 30 seconds with audio. Per 10 seconds, that price is real and matchable — it is exactly what the cheapest H3 route sustains. Per 30 seconds, it is not fundable from any published API rate we could find: the advertised model, Seedance 2.5, costs $0.231 per second at 720P on its official API, which is $6.93 of generation alone for 30 seconds. A $0.45 price for that output has to be subsidized from somewhere else. When you compare offers, compare the same unit: 10 seconds, a stated resolution, audio on or off.
There is also a capability catch with that particular model for film work: Seedance refuses reference images containing realistic human faces, which rules it out for recurring photoreal characters — the workflow in our character consistency guide. Cheap per second is only cheap if the model accepts the inputs your film needs.
We keep it deliberately boring: Draft videos are $0.59 and Final videos are $1.15, per 10-second clip at 768P, generated on the MiniMax H3 family — Drafts on the Lite tier, Finals on the official API route. No subscription, and failed generations are never charged. The full pack list is on our pricing page. NascoreAI is currently in pre-launch — the studio runs in demo mode while payments are activated — and these are the launch prices.