The MiniMax model family
Multimodal video generation with native stereo audio.
Multimodal video generation with native stereo audio.
Generation is not connected yet. Explore the controls and prepare your next idea.
Model names and parameters share the site catalog. Availability depends on a valid model connection, the selected settings, and account access. A displayed model is not a successful API test.
Compare compatible inputs and output settings before starting your draft.
Add a clear prompt and reference material when the selected model supports it.
Inspect completed outputs and download the version that fits your project.
Multimodal video generation with native stereo audio.
Start with references, aspect ratio and output resolution. Video tasks also need a duration, sound and frame-input check. These settings share the same server-side model catalog.
Compare specifications and workflows across the same media type, then prepare your own direction with a consistent prompt. The catalog is not a provider connectivity check or a performance ranking.
These original visual examples illustrate creative directions. They are not results generated by the currently unconfigured website service.
Text, image and reference video generation with native audio.
Accelerated text and image video generation.
Long-take text and image video with synchronized audio.
Reference-site fast model label; a matching provider endpoint is required.
Faster video drafts from text or a first frame.
Text and image video with native audio.
Plan preview. Purchasing is not open yet.
Billed annually, total: $120.00
Billed annually, total: $240.00
Billed annually, total: $480.00
Annual plans bill the full year and grant annual credits upfront. Purchasing is closed. No payment will be taken. Refund Policy
Answers about source files, model settings, credits, task history and responsible use.
MiniMax H3 belongs to the MiniMax model family. Multimodal video generation with native stereo audio. This page describes the CreoArt model configuration; it does not indicate that API credentials are already connected.
Write a prompt, choose a model and settings, then review the estimate before submitting.
You can explore the workspace and prepare a local draft before signing in. Submitting and managing account-owned generation tasks requires an account.
Video workflows accept the file types and duration limits displayed beside the uploader. Text-to-video starts from a prompt; image-to-video starts from an image.
Describe the subject, action, setting and camera direction. Mention details that must remain consistent across the shot.
Configured ratios: 16:9, 9:16, 1:1, 4:3, 3:4, 21:9; resolutions: 480P, 768P, 2K, 4K. Final availability is checked against the server configuration and provider contract.
Credits depend on the model, options and output count. The server calculates the applicable estimate before submission; a client-entered amount is not authoritative.
No. Local editing controls help you prepare a draft, but a model connection must be configured before an inference request can run. The site must not label a preview as a generated result.
Resolution options come from the selected model. Not every model supports every size; the controls and model detail page show the configured capability range.
Choose a supported ratio in the task settings. For edits and expansions, also check the original framing and the target canvas before submitting.
Timing depends on the provider, selected model, options and queue. Follow the task status rather than assuming a guaranteed completion time.
Refine the instruction, simplify conflicting requirements or choose a different compatible model. A new generation can incur a new cost, so check the estimate again.
Open My creations while signed in to review account-owned history. A completed task and a durable downloadable asset are separate states; follow the result and retention information shown.
Usage depends on the model provider terms, the inputs you own or license, and the applicable product terms. A paid plan alone is not a blanket copyright guarantee.
A local file preview stays in the browser until an upload or task is submitted. Connected generation may send the necessary material to the configured provider; review the Privacy Policy.
Use text-to-video for an idea without a source frame, image-to-video to animate a picture, video editing to change an existing clip, and upscaling to improve delivery resolution.