Generate MiniMax H3 Max video from text, images, or frames

MiniMax H3 Max is a post-trained MiniMax H3 video model. It follows prompts more tightly and finishes shots faster, while keeping the same visual character. This page is the working studio: text to video, image to video, and first-last frame, with duration, resolution, and aspect ratio on one screen.

Use it for product loops, social ads, fashion motion, concept-art tests, and animatics. Clips last 5 to 15 seconds. Text-to-video can pick 21:9, 16:9, 4:3, 1:1, 3:4, or 9:16. Image-led jobs follow the still you upload. 480P costs 1 credit per second; 768P costs 2 credits per second.

h3maxai.com is an independent studio, not the official MiniMax homepage. Sign in with Google, buy credits with Stripe, and generation reserves credits first. Failed tasks return reserved credits so you can revise the prompt or frames and try again.

Text to video invents the shot from language. Write the camera, lighting, performance, wardrobe, location, and motion. Start here when you have no stills, or when you want a new composition rather than animating a photograph.

Image to video keeps the still you upload as the first frame. The model reads composition, subject, and palette, then adds motion from the prompt: a dolly-in, hair in wind, a product orbit, rain on glass. Identity stays close to the photograph. Aspect ratio follows the image.

First and last frame generation takes two stills. The first image is the opening composition; the last image is the destination. The prompt describes the journey between them — weather, time of day, travel, a costume change. The in-between is interpolated so the shot can leave a rainy street and arrive on a ridgeline without cutting.

Prompt expansion has two speeds. Balanced rewrites the prompt in about a second. Quality spends up to ~30 seconds on a richer prompt before rendering. Neither mode changes the credit cost. Credits scale only with duration and resolution: 480P is 1 credit per second, 768P is 2.

Output is 480P or 768P. 480P is the cheaper preview and social default. 768P is the sharper finish. Aspect ratios cover cinematic widescreen, social vertical, square, and classic photo shapes on text-to-video. Image-led jobs inherit ratio from the source still.

H3 Max resolution: 480P vs 768P

Duration runs from 5 through 15 seconds. Longer shots cost more because they contain more frames. A 5-second 480P clip costs 5 credits; the same length at 768P costs 10. Write camera and motion clearly in the prompt. This studio does not take reference video or audio.

Uploaded stills should stay between 256 and 5760 pixels on each side, with aspect ratios between 0.4 and 2.5. JPEG, PNG, WebP, and HEIC are accepted. Those limits keep jobs inside the allowed file range so a still is less likely to fail on size.

Pick an example in the gallery, or write your own prompt above. Examples load the prompt and images for you. Read the prompt before you generate — a loaded case is a starting point, not a locked template. You can swap a still, shorten the duration, or switch from 480P to 768P without rewriting the whole shot.

Sign in with Google, then open pricing to buy a monthly plan or a one-time credit pack. Starter, Creator, and Studio subscriptions grant video credits each month. Packs add credits without a subscription. Generation reserves credits first; a failed run returns them.

Set duration, resolution, and aspect ratio, then generate. The preview panel shows progress and the finished clip. History lives under My videos after you sign in. Iterate by changing one variable at a time — prompt wording, a still, or 480P versus 768P.