MiniMax H3 costs $0.1189 per second at 2K on HiAPI. H3 Max costs $0.072 per second at 480P or $0.115 at 768P, so a 10-second clip is $1.189, $0.72 or $1.15. Max 768P is slightly cheaper per second than H3, but it outputs a different resolution and takes different inputs.
First test: a 5-second paper-boat clip at Max 768P has a $0.575 generation budget. Generate the paper-boat test →, then select H3 Max, 5 seconds, 768P and 16:9 manually; the link fills only the prompt.
What does a 5-, 10-, or 15-second clip cost?
Budget = per-second rate × clip length. H3 accepts integer durations from 4 to 15 seconds; H3 Max accepts 5 to 15 seconds. Both default to 5 seconds, so set the duration explicitly when planning a batch.
| Length | H3 2K | H3 Max 480P | H3 Max 768P |
|---|---|---|---|
| Per second | $0.1189 | $0.072 | $0.115 |
| 5 seconds | $0.5945 | $0.36 | $0.575 |
| 10 seconds | $1.189 | $0.72 | $1.15 |
| 15 seconds | $1.7835 | $1.08 | $1.725 |
A short product movement or a single camera move is a useful starting point for a 5-second request. Choose 10 or 15 seconds when the shot needs room for an opening, a visible change, and a finish. Decide what happens during those extra seconds before extending the request.
How many clips fit into a $10 generation budget?
At 5 seconds each, $10 covers 16 H3 2K clips, 27 Max 480P clips, or 17 Max 768P clips when rounded down to whole requests. Those whole clips total 80, 135 and 85 seconds; other clip lengths change the count.
For a first test, generate with this prompt →. Choose H3 Max, 768P, 5 seconds and 16:9: $0.575 per clip. The link fills in the prompt; choose the other settings in the playground.
This clip was requested with H3 Max, 768P, 5 seconds and 16:9; the output is 1344×768. The three-sentence prompt covers the subject, camera path, water movement and ambient sound:
After a morning rain, a red paper boat drifts through shallow water beneath a stone bridge. A low camera tracks smoothly beside it, small ripples spread across the water, and birds sing softly in the distance. No text or subtitles.
Start by checking whether the camera follows the subject as intended. Once the motion fits your edit, choose the duration that the final scene needs. Keep the prompt focused on one movement so each candidate answers a clear creative question.
What should you budget for a monthly batch?
Estimate total seconds before choosing the resolution. Here is a budget matrix for 10-second clips; it covers video generation, with editing and storage handled separately.
| Monthly volume | H3 2K | H3 Max 480P | H3 Max 768P |
|---|---|---|---|
| 10 clips / 100 seconds | $11.89 | $7.20 | $11.50 |
| 100 clips / 1,000 seconds | $118.90 | $72 | $115 |
| 1,000 clips / 10,000 seconds | $1,189 | $720 | $1,150 |
Allocate candidate shots separately from selected shots. A sequence that will use one clip may still need several generation attempts while you settle the action or framing. Failed tasks are refunded; include successfully generated candidates in your budget even when they do not reach the final edit.
Generate the paper-boat test → before planning a batch. Select 5 seconds, 768P and 16:9 manually.
Which model makes sense for your budget?
Pick Max 480P for landscape drafts; choose H3 when you need 2K output or image, video, and audio references. Max 768P fits work that needs a higher output resolution than 480P while retaining the Max first- and last-frame workflow.
| Decision | H3 | H3 Max |
|---|---|---|
| Output resolution | 2K | 480P / 768P |
| Minimum duration | 4 seconds | 5 seconds |
| Frame control | Last frame requires a first frame | First or last frame independently, or both |
| Reference inputs | Up to 5 images, 3 videos, 3 audio clips | First- and last-frame inputs |
| Text-only portrait work | 9:16 is available | Text-only output is currently landscape; sample: 1344×768 |
H3 Max text-only generation currently produces landscape output: the new 768P sample is 1344×768. For portrait delivery, use a portrait first-frame image and omit aspect_ratio, or crop or pad the output in editing.
Max 768P costs $0.0039 less per second than H3, about 3.3%. That price difference comes with different output resolutions and supported inputs. Choose the workflow that can use your assets and deliver your required resolution before comparing the rates.
Within Max, the 768P rate is about 1.6 times the 480P rate. Use 480P to settle the direction and 768P for shots that need the higher resolution. Repeated drafts still add up, so reduce the candidate list before extending every shot.
For H3 reference work, frame control and the image/video/audio reference lists are separate input modes. Audio references accompany an image or video. Moving an existing H3 reference workflow to Max therefore requires changing its inputs, beyond changing the model ID.
Pricing FAQ
Is H3 Max a higher-resolution H3 tier?
They are separate models. H3 delivers 2K; Max offers 480P and 768P. The Max name does not imply a higher price or a higher output resolution.
Can I request 4 seconds with either model?
H3 accepts 4 seconds, which costs $0.4756 at 2K. Start Max requests at 5 seconds and choose any integer duration through 15 seconds.
Which resolution strings should I send?
Use 2K for H3 and 480P or 768P for Max. Preserve the uppercase letters and send an explicit integer duration so each request matches the budget you planned.
Generate the paper-boat shot with this prompt →, select 5 seconds, 768P and 16:9 manually, then choose the final length after checking its motion. For more reusable shots, see the H3 and H3 Max prompt guide; for input fields and integration details, read the H3 Max API documentation.
