Skier facing the camera
A skier pauses on a snowy slope and looks toward the camera as powder drifts through cinematic backlight and the camera pushes in.
Save with annual plans 90% OFF All Models · Ends Oct 31Annual plan savings 90% OFF All Models · Ends Oct 31 (Yearly Plans)
Create MiniMax H3 videos from text and images on h3ai.org. Explore free starter credits, prompt examples, max duration and member pricing in H3 AI.
A skier pauses on a snowy slope and looks toward the camera as powder drifts through cinematic backlight and the camera pushes in.
A female warrior in futuristic armor walks through a neon city in the rain, tracked by a dynamic camera.
An anime heroine stands in a windswept meadow, hair and clothes moving naturally in soft sunset light.
A giant surreal eye floats above a waterfall canyon while clouds pass through an otherworldly cinematic frame.
A macro camera glides over glowing circuits as energy races across a chip in blue and violet light.
An old book opens and its pages turn into birds flying toward a golden sky through dreamlike particles and volumetric light.
A man in futuristic clothing drinks tea in a minimal space as the camera circles through delicate steam and soft light.
An anime coastal train passes beneath a sunset while clouds drift slowly through warm film colors.

Compare the visual language, motion, and camera control of available models for cinematic, animated, and product work.

Choose only the ratios, durations, and resolutions the selected model actually supports, with every setting visible before submission.

Paid workflows include private creation and supported watermark-free results; commercial use still depends on model terms and source rights.

Keep prompts and results together, compare drafts across models, and spend credits on the version that fits best.
Compare MiniMax H3 credits and member prices on H3 AI. Explore free starter credits and Unlimited image benefits; H3 videos always use credits.
Cancel anytime • Unused credits roll over
MiniMax H3 is an AI video model for text-to-video, image-to-video, first and last frames, and multimodal reference workflows. H3 AI provides browser-based access on h3ai.org.
Keep MiniMax H3 selected in the generator, choose text-to-video or add reference media, describe the shot, camera, motion and sound, review the supported settings, then generate.
MiniMax H3 generation on H3 AI supports 4 to 15 second clips. The duration options shown in the generator are authoritative for the selected workflow and plan.
On H3 AI, the AutoDL workflows offer 480P and 768P, plus 1080P for supported image-reference clips of up to 10 seconds. The official model also has a 2K path, which is not offered by this integration.
H3 AI uses hosted AutoDL workflows. For MiniMax H3 ComfyUI or local setup, check the official repository and verify the node version, model license and GPU requirements; a hosted API is not a local model download.
MiniMax publishes H3 resources on its official GitHub. For model downloads, local deployment and GPU requirements, follow that repository. H3 AI runs in your browser and does not install model weights on your device.
H3 AI applies safety and content rules to prompts, uploads, and outputs. Sexually explicit, exploitative, illegal, or otherwise prohibited content is not supported.
New accounts can use any starter credits currently offered to test supported workflows. Access is limited rather than permanently free, and cost depends on the selected model and settings.
Unlimited plans include zero-credit generation for eligible image models. Every video model consumes credits, with 50% off on monthly Unlimited and 60% off on yearly Unlimited; SOTA image models also use credits.
H3 AI does not claim ownership of your generated output. Commercial use depends on your plan, the selected model's terms, and your rights to prompts and uploaded assets; you remain responsible for lawful use of the final content.
H3 AI supports professional workflows for marketing, social content, product visuals, client drafts, and production tests. Paid plans add private generation, watermark-free supported outputs, higher concurrency, and commercial-use workflows as listed on the pricing page.
Generation time varies by model, mode, duration, resolution, references, queue demand, and provider status. The workspace shows the current task state while your result is being created.
Start with a prompt, first frame, last frame, or reference media, then choose supported H3 settings in one workspace.