AI API Prices › Video generation

AI video generation API pricing

Official cost per generated second for MiniMax H3, Seedance 2.5, and Wan 3.0. Resolution tiers and reference-video billing are shown explicitly, because one headline number is not enough for an accurate estimate.

Image API pricing LLM token pricing

USD list prices

Representative text/image-to-video output rate, with no input video. Prices verified 2026-09-13; use each row’s official source for budgeting.

ModelProviderFrom / secondReference tierMax duration
MiniMax H3
Hosted API with native stereo audio; open weights generate at native 768p and self-hosting has separate compute cost
 MiniMax $0.0800 768p output 15s
Seedance 2.5
Native audio and multimodal references; Seedance 2.5 does not support 1080p or 4K in this official API route
 ByteDance / BytePlus $0.2056 480p · no input video 30s

Wan 3.0 native CNY list prices

Alibaba Cloud’s official list price is CNY per second. No USD conversion is applied.

Model480p720p1080pBilling
Wan 3.0 Video
API: wan3.0-video
 ¥0.3000/s¥0.6000/s¥1.20/s Text-to-video bills successful output seconds. When an input video is supplied, billable duration is input video duration plus successful output duration.
Wan 3.0 Video Prime
API: wan3.0-video-prime
 ¥0.4500/s¥0.9000/s¥1.80/s Text-to-video bills successful output seconds. When an input video is supplied, billable duration is input video duration plus successful output duration.

Price tiers and official sources

MiniMax H3

Generated video is billed per output second at the selected resolution.

768p: $0.0800/s2K: $0.1300/s

Additional billing: Audio inputs are free. The first 5 images are free, then $0.04 per additional image. Input video is billed by its duration at the selected output-resolution rate. 768p-to-2K regeneration is $0.05 per output second; regeneration input uses $0.05 per base-video second and $0.025 per image after the first 5. H3-Context-IR is $0.90/M input tokens and $3.60/M output tokens.

Official pricing ↗ · verified 2026-09-13

Seedance 2.5

For generation without an input video, effective price is the official $0.303 enhanced-tier base rate × resolution coefficient × output seconds (480p: 0.6785; 720p: 1.525).

480p · no input video: $0.2056/s720p · no input video: $0.4621/s

Additional billing: With input video, both input and output duration are billed using different coefficients: 0.406 at 480p or 0.9125 at 720p, multiplied by the $0.303 base rate.

Official pricing ↗ · verified 2026-09-13

Wan 3.0 Video

Text-to-video bills successful output seconds. When an input video is supplied, billable duration is input video duration plus successful output duration.

480p: ¥0.3000/s720p: ¥0.6000/s1080p: ¥1.20/s

Singapore international deployment: ¥0.37471/s (480p), ¥0.74942/s (720p), ¥1.49884/s (1080p).

Official pricing ↗ · verified 2026-09-13

Wan 3.0 Video Prime

Text-to-video bills successful output seconds. When an input video is supplied, billable duration is input video duration plus successful output duration.

480p: ¥0.4500/s720p: ¥0.9000/s1080p: ¥1.80/s

Singapore international deployment: ¥0.495838/s (480p), ¥1.020844/s (720p), ¥2.041687/s (1080p).

Official pricing ↗ · verified 2026-09-13

Video pricing FAQ

How are AI video APIs priced?

These APIs bill successful generation by video duration, usually per second, with higher rates for higher output resolutions. Reference-video workflows may also bill input duration or add material fees.

Why are Wan 3.0 prices shown in CNY?

Alibaba Cloud publishes these list prices in Chinese yuan. Keeping the native currency is more accurate than presenting a fixed USD conversion that changes with the exchange rate.

Are input videos billed?

It depends on the provider. Wan 3.0 adds input-video and output-video duration. MiniMax H3 bills input video at the chosen output-resolution rate. Seedance 2.5 uses separate input-video coefficients. Text-to-video has no input-video charge.