MiniMax API models and pricing

How to choose a MiniMax model

Start with the material you have. A text-only request describes the subject, action, camera and setting. A first-frame workflow starts from an image; first-and-last-frame and reference modes offer different controls when the selected model supports them. Check the table before uploading references, since model variants do not accept identical inputs.

Then choose the clip length and resolution. For a sequence of shots, test a short clip first so you can judge movement, composition and continuity before spending on longer output. Sound support and the available aspect ratios also vary by model. Use the live catalog for the complete request contract.

Quote and generate through EveryFrame

Create an API key in the console. Keep it on your server and use the API guide to quote a request, submit a render, poll for completion and retrieve the finished media. The quote reflects the chosen inputs and current API price; API funding is separate from Studio credits.

Compare with other video models

Choosing by name alone misses differences in input controls, resolution, length and sound. Compare the same scene with the same settings, and assess usable motion and consistency as well as price. Explore Seedance · Kling · Veo · MiniMax or use the comparison view in the full model catalog.

Get an API key Try in Studio

See a real MiniMax H3 Max result

Manta rays above a coral reef

15-second gallery video, made with MiniMax H3 Max. The watch page includes playback and the original prompt.

Two giant manta rays glide in slow graceful loops over a coral reef, sunbeams dancing through the blue water as schools of silver fish scatter around them. Underwater documentary, wide-angle lens, natural light rays, rich turquoise palette. Sound: muffled ocean ambience, gentle bubbles, a serene orchestral swell,…

Watch the output and read the full prompt ↗

This example illustrates one model and prompt. It is not a controlled comparison with other models, and the current price examples above do not represent its historical cost.

How Bittensor fits into EveryFrame

EveryFrame combines its creative studio and API with SN117, a network where miners serve supported model requests through provider accounts. Learn how Bittensor AI video generation works.

MiniMax H3 Max: input limits and a working request shape

The limits below describe h3-max-fast in the EveryFrame catalog. They are not a promise that every MiniMax variant or a direct provider account has the same features. Choose the exact model ID before building an upload form.

File ceilings are upper bounds, not a requirement to supply that many files. A first-frame request uses its opening image; other reference modes have their own validation. Text-only requests need no reference upload. Read the upload documentation before sending media, and keep the catalog response alongside your model picker so unsupported combinations never reach your users.

Quote a 5-second 768p clip

Set EVERYFRAME_API_KEY in your server environment. This example asks for a price and does not start a paid generation. Inspect the returned resolution, duration and USD amount before submitting the same render settings to POST /v1/render.

curl https://api.everyframe.studio/v1/render/quote \
  -H "Authorization: Bearer $EVERYFRAME_API_KEY" \
  -H 'Content-Type: application/json' \
  -d '{
  "modelId": "h3-max-fast",
  "mode": "video",
  "prompt": "A paper lantern drifts above a quiet lake at dusk, slow tracking shot",
  "duration": 5,
  "resolution": "768p",
  "aspect": "16:9",
  "referenceMode": "text"
}'

After submitting a render, retain its ID and poll its status using the API guide. A submission response is not a finished video. Download only after completion, and show failed requests clearly instead of leaving your own interface in an endless loading state. Keep your API key out of browser code and shared examples.

EveryFrame API or direct MiniMax integration?

Choose a direct integration when you require a provider-specific control that EveryFrame does not expose. Choose a shared API when switching models and managing one application workflow matters more. Neither option guarantees a better result for every scene. Compare usable outputs for your actual prompts, include failed experiments in your budget, and avoid comparing a per-second rate with a per-clip total.