Seven reference images in a 5-second MiniMax H3 clip: $0.52 at 480p
MiniMax H3 reference-to-video includes five reference images and charges $0.08 list for each extra. A 5-second 480p clip with seven images is $0.52 on Sume.
A 5-second MiniMax H3 clip at 480p with seven reference images costs $0.52 on Sume: the base clip, then two extra images at $0.08 list each, all times 1.25. With five images or fewer the same clip is $0.32.
The arithmetic
Sume's catalog entry for minimax-h3 says that in reference-to-video the first five reference images are free and each further image adds $0.08 list. The base list is $0.05 per second at 480p, so 5 seconds is $0.25. Two extra images add $0.16 list. The sum, $0.41, is multiplied by 1.25 to $0.5125 and rounded up to the cent.
| Images | Extra images | List total | Billable |
|---|---|---|---|
| 1 | 0 | $0.25 | $0.32 |
| 2 | 0 | $0.25 | $0.32 |
| 3 | 0 | $0.25 | $0.32 |
| 4 | 0 | $0.25 | $0.32 |
| 5 | 0 | $0.25 | $0.32 |
| 6 | 1 | $0.33 | $0.42 |
| 7 | 2 | $0.41 | $0.52 |
| 8 | 3 | $0.49 | $0.62 |
| 9 | 4 | $0.57 | $0.72 |
What the extra images buy
Each reference image can fix one thing: a character, a product side, a location, a garment. Seven images means two more than the free five. If two of them show the same object from similar angles, drop them and keep the price at $0.32.
The cap is nine images, and images, videos and audio clips together may not exceed twelve. Audio cannot be the only reference. Up to three reference videos and three audio clips are accepted, each 2 to 15 seconds and 15 seconds combined.
- Up to 9 reference images.
- Up to 3 reference videos and 3 audio clips, combined with images no more than 12.
- The surcharge applies to MiniMax H3; MiniMax H3 Max bills r2v as output seconds only.
Cheaper ways to the same shot
If you need more than five images, check whether H3 Max at 480p fits. It has the same $0.05 list per second at 480p and no per-image fee, so a 5-second, seven-image clip there is $0.32 against $0.52 on H3.
Practical advice for reference sets
Choose reference images by role, not by count. A typical set for a product clip is one hero shot, one close-up of the material, one side or back view, one person using it and one location. That is five images and costs nothing extra. A sixth and seventh image, such as a logo card and a packaging shot, are worth paying for only if the clip must show them. Image quality matters more than quantity: sharp, evenly lit images on plain backgrounds give the model less to guess. Refer to each image in the prompt by its order, and keep the prompt short enough that each reference has a clear job.
Whichever you pick, count the images before you submit. A request with more than nine reference images is refused, and a request with audio as its only reference is refused as well. Stay within twelve files in total across images, videos and audio clips.
Sources
Related posts
More in Pricing
- Shopify plays video at 480p to 1080p: what a 4K source costs on Sume
Shopify says streaming playback adjusts between 480p, 720p and 1080p though uploads reach 4K. A 10-second Omni clip is $1.88 at 1080p and $3.75 at 4K on Sume.
- Shotstack renders at 5 to 20 cents a minute; Sume Timeline is 10
Shotstack lists 20 cents a minute pay as you go down to 5 cents on Volume. Sume Timeline is a flat $0.10 per output minute. Where the lines cross.
- Mix 480p and 720p Seedance 2.5 clips in one 180 s Short: $67.26
Four 480p and two 720p Seedance 2.5 clips of 30 s make a 180 s Short for $67.26 on Sume. Drafting all six at 480p first costs $83.40. The cost table is inside.
- Sonix $10 an hour vs Sume's $0.01-a-minute transcript
Sonix Pay As You Go is $10 an hour, about 16.7 cents a minute. Sume's transcript line is $0.01 per audio minute, plus inspect compute. Compared fairly.
Written by Sume