MiniMax: MiniMax H3 Max
minimax-h3-max- Price
- From $0.08
- Modalities
- Video
About MiniMax H3 Max
MiniMax H3 Max is a variant of the open-weight MiniMax H3 video model, post-trained by fal Research for prompt adherence and aesthetics and served on fal's own inference stack. It generates video with natively synchronized audio from a text prompt. fal says it renders a short clip many times faster than the original H3 endpoint, and it is offered for text-to-video.
Where it works well
- Audio is generated together with the picture, so a clip arrives with sound that matches the action instead of needing a separate sound step.
- fal reports a five-second 768p clip in about 2.8 seconds, against about 53 seconds for original H3, which makes iterating on prompts quick.
- Post-training targets prompt following, and fal reports first place in its own human-preference tests plus Artificial Analysis and Design Arena rankings.
- The request is simple: a text prompt and fixed output settings, with the result retrieved through asynchronous task polling.
When to choose another model
- Output is fixed at five seconds, 768p and 16:9, so longer clips or 1080p need another model.
- It is text-to-video only, so animating a still image needs a different model.
- The model is fal's modified version rather than MiniMax's own release, so outputs differ from the original H3 and its documentation.
Getting started
Create API key
Create a key in Console, then use it with every model on the platform.
Send your first request
Copy the example for your language and run it against the endpoint.
Text to videoTokenLab endpointPOST/v1/videos/generationscurl -X POST "https://api.tokenlab.sh/v1/videos/generations" \ -H "Authorization: Bearer sk-xxx" \ -H "Content-Type: application/json" \ -d '{ "aspect_ratio": "16:9", "duration": 5, "operation": "text-to-video", "resolution": "768p", "model": "minimax-h3-max", "prompt": "Cinematic sunrise over a calm lake with gentle camera motion." }'
Pricing
The Verified price applies to Verified, which costs less on most models. The Official price is the model maker's published price and applies to the more reliable Official route. Auto bills the route that completes the request.
Text to video
per second- Official price
- $0.08
- Verified price
- —
- Discount
- —
| Official priceper second | Verified priceper second | Discount | |
|---|---|---|---|
| Text to video | $0.08 | — | — |
Usage & activity
Success rate is the share of requests that completed. Latency is how long a full response takes; P95 means 95% of requests finished within that time.
Usage & availability
Last 24 hours- Requests
- Success rate
- P95 latency
- Total tokens
Data is based on aggregate user requests, excluding status checks.
Open in Console
Open MiniMax H3 Max in Console with a prompt ready to edit or send.
Help me create with minimax-h3-max using text-to-video at /v1/videos/generations. Show the result, status, and cost.
Use cases
Best for- Text to Video
Rapid concept drafts
Test many versions of a scene description in minutes and keep the ones where composition and motion read well.
Short clips with sound
Make social or product teaser shots that include ambient sound, voices or effects without an editing pass.
Prompt adherence testing
Check how closely a model follows detailed instructions about camera, subject action and mood before committing a larger budget.
Storyboard animatics
Turn scripted shots into moving, sounded previews so a team can judge pacing before production.
Prompt examples
A street vendor flips noodles in a wok at night, steam rising through neon light, handheld camera slowly pushing in, sizzling and city chatter.
A glass of iced coffee on a sunlit windowsill, condensation sliding down, a gentle breeze moving the curtain, soft piano in the background.
Two astronauts walk across a red desert toward a distant base; wide tracking shot, wind noise and their muffled breathing.
FAQ
How is MiniMax H3 Max different from MiniMax H3?
H3 Max is fal's post-trained variant of the open-weight MiniMax H3. fal says it follows prompts better, looks better, and runs on an inference stack co-designed with the model for speed. The weights and behavior are not identical to MiniMax's own H3 release.
Does MiniMax H3 Max generate audio?
Yes. Video and audio are produced together, so speech, effects and ambience are tied to what happens on screen. Describe the sounds you want in the prompt, such as dialogue, wind or music, because the model uses the text to decide them.
What clip length and resolution can I get?
One combination: five seconds at 768p in 16:9. Longer clips or 1080p need another model. Because the output settings are fixed, you only write the prompt and receive a short clip that is quick to review.
Can I start from an image with MiniMax H3 Max?
No. It generates from a text prompt only. To animate a still frame, choose a model that offers image-to-video; this one is for prompts written from scratch, so describe the subject, camera and sound in text.
How do results arrive?
Generation runs as an asynchronous task. You submit the prompt, receive a task identifier, and poll until the video is ready. Because fal reports generation in seconds, polling usually finishes quickly, but the pattern is the same as for slower video models.
How much does MiniMax H3 Max cost?
On TokenLab, MiniMax H3 Max costs $0.08 per second. The pricing table above shows the full breakdown. Rates depend on the billing unit, specification, and usage. Compare matching conditions in the model's detailed pricing; a single rate does not determine the total cost.
Which endpoint should MiniMax H3 Max use?
Use https://api.tokenlab.sh/v1/videos/generations for MiniMax H3 Max. The request example below shows the matching code shape.
Which operations does MiniMax H3 Max support?
MiniMax H3 Max supports Text to video. Select an operation above to see its endpoint and request example.
Compare MiniMax H3 Max
Sources
Reviewed Oct 2, 2026
