MiniMax H3 Max reference video: 5 seconds of input adds 71 cents
MiniMax bills H3 Max reference video at $0.143 per second at 768P, on top of output. A 10-second clip with 5 seconds of reference is $1.515 direct.

On MiniMax's page, H3 Max reference video is billed by input duration at $0.143 per second at 768P, so 5 seconds of reference video adds $0.715. A 10-second 768P H3 Max clip with that reference therefore costs 10 x $0.08 + 5 x $0.143 = $0.80 + $0.715 = $1.515. Reference audio is free, and images are free up to 2 per request, then $0.074 each.
| Model | Audio | Image | Reference video |
|---|---|---|---|
| MiniMax-H3 | Free | First 5 free, then $0.04 each | 768P $0.08 per second; 2K $0.13 per second |
| MiniMax-H3-Max | Free | First 2 free, then $0.074 each | 480P $0.0553 per second; 768P $0.143 per second |
| MiniMax-H3-Regeneration | Free | First 5 free, then $0.025 each | $0.05 per second of the original 768P task |
Runway and Sume
Runway's page bills hailuo3 reference video at the same per-second rate as output, capped at 15 seconds, plus 2 credits per reference image or keyframe. At 768P that is 10 credits per second, so 5 seconds of reference video costs 50 credits, or $0.50, on top of the output.
Sume's docs say minimax-h3-max accepts image, video and audio references. The catalog prices it per output second by resolution ($0.0625 at 480p, $0.10 at 768p, $0.20 at 1080p), and the rows we read show no separate input-material line, so we do not claim a reference-video fee on Sume either way. Check the live estimate from GET /v1/videos/models before a large job.
Planning a reference-heavy job
Reference video is the expensive input. On MiniMax direct, the 5-second reference in the example costs 0.715 / 1.515 = 47 percent of the whole bill. Trimming the reference to the 2 seconds you need drops its charge to 2 x $0.143 = $0.286. Sume's Video trim tool costs $0.02 per job, so trimming first is cheap on any platform.
Images versus video
Reference images are much cheaper than reference video on the MiniMax page. Four extra images on H3 Max beyond the 2 free cost 4 x $0.074 = $0.296, less than half the $0.715 for 5 seconds of reference video at 768P. If a still can carry the idea, use it; reserve video references for motion you cannot describe.
The H3 row for comparison
The standard H3 row charges reference video at the output rate for its resolution: $0.08 per second at 768P and $0.13 at 2K. Five seconds of 768P reference therefore adds 5 x $0.08 = $0.40 on H3 against $0.715 on H3 Max, even though H3 Max's output rate is the same $0.08 at 768P. If your job allows either model, the reference-video line is where H3 is cheaper.
Check before you submit
MiniMax's page is the source for every number in this post, and the Runway page for the Runway figures. Confirm them again before a large batch, since both can change. On Sume, read the live estimate for your exact request before you send it.
Sources
Related posts
More in Pricing
- MiniMax H3 with 7 reference images and 6 s at 768p: $0.65 on Sume
On MiniMax H3 the first 5 reference images are free and each extra one adds $0.10 on Sume. Seven images and 6 seconds at 768p come to $0.45 + $0.20 = $0.65.
- Music for 21 scenes: 1 to 5 takes each at $0.125 flat on Sume
Scoring 21 scenes costs $2.63 with one take each and $13.13 with five on Sume Music 1.0, which charges a flat $0.125 per generation.
- Nano Banana 2.1, 1,000 images: Sume row vs Google's standard and batch
1,000 Nano Banana 2.1 images at 2K: $150 on Sume's row, $50.40 at Google's standard rate, $25.20 at Google's batch rate, per each page read 2026-10-09.
- Nano Banana 2.1 for 37 images: Google's price vs Sume at 1K, 2K, 4K
A 37-image set on Nano Banana 2.1 costs $1.24 to $4.18 at Google's list and $3.70 to $7.40 on Sume. Per-tier arithmetic and what the gap buys you.
Written by Sume