MiniMax Hailuo H3 Max Image to Video

MiniMax Hailuo H3 Max Image to Video

MiniMax Hailuo H3 Max Image to Video is the top-quality tier of the H3 (Hailuo-03) family, turning a still image into a moving clip with native stereo audio.

What is MiniMax Hailuo H3 Max Image to Video?

H3 is MiniMax's general-purpose, multimodal video model, and Max is its highest-fidelity tier, tuned for the hardest prompts where lighter tiers lose coherence. On this image-to-video endpoint you supply a first frame (and, optionally, a last frame) along with a text prompt describing the motion. The model animates the still at 480P or 768P while keeping its composition and subjects intact, so the frame you start from is the frame that moves. Provide both a first and last frame and the clip interpolates cleanly between the two.

Key Features

• •Image-to-video from a first frame, a last frame, or both (first-to-last interpolation)
• •Text prompt up to 7000 characters to direct the motion
• •Selectable resolution: 480P or 768P
• •Clip duration from 5 to 15 seconds
• •Native stereo audio generated alongside the video
• •Prompt-expansion modes: disabled, balanced, or quality
• •Max tier for maximum fidelity on complex, physically demanding motion

Best Use Cases

Reach for H3 Max image-to-video when a still needs demanding motion added: fluids and particulates, fast or articulated action, reflective or transparent materials, and camera moves motivated by the scene. In testing it animated a cryogenics-lab still into dense liquid-nitrogen fog that billowed from the flask, cascaded over the bench edge and rolled low across the floor while the row of glass beakers it swallowed stayed sharp and correctly reflective — complex fluid motion that held together instead of smearing. It suits animating product and science stills, turning concept art or key frames into finished shots, building first-to-last-frame transitions, and social clips where one legible action must read at a glance.

Prompt Tips and Output Quality

Compose the first frame so the element that should move is already present in it. Lead the prompt with one clear primary action, then add supporting detail. Keep prompt-expansion on balanced for a well-written prompt and switch to quality to enrich a sparse one. Use 768P for finals and 480P for quick drafts, and give longer durations to actions that need room to play out.

How to Use MiniMax Hailuo H3 Max Image to Video

MiniMax Hailuo H3 Max Image to Video animates a still image into a short clip with native stereo audio. This guide covers how to prepare your input frame, write the motion prompt, and pick the right settings for each job.

Start with the right frame

The model keeps your frame's composition and subjects and adds the motion you describe, so the input frame does most of the work. Compose it so the element that should move is already in shot — a liquid about to pour, a vehicle on an open road, fabric in a breeze. Supply a first_frame_image to drive the clip from its opening still, or add a last_frame_image as well: with both frames the model interpolates between them for a defined start-to-end motion. The output's aspect ratio follows the frame you provide.

Write the motion prompt

Describe the subject and, above all, the motion to add. Lead with one clear primary action and keep secondary beats to a minimum — dense multi-action prompts can drop a detail. Prompts can run up to 7000 characters, but a tight, concrete description reads best. Leave prompt_expansion_mode on balanced for a well-written prompt; switch to quality to enrich a sparse one, or disabled to keep your wording verbatim.

Recommended Settings

• •resolution — the main quality lever. Use 768P for finished clips and 480P as a fast, economical draft pass.
• •duration — 5 to 15 seconds (default 6). Six to eight seconds lets a single action play out cleanly; longer clips give complex motion room to resolve.
• •prompt_expansion_mode — balanced for detailed prompts, quality for sparse ones, disabled to preserve exact wording.

In testing, a 768P, six-second clip animated a cryogenics-lab still into dense liquid-nitrogen fog that cascaded over the bench while the glass beakers stayed sharp and reflective — a clean, artifact-free result with synced stereo audio and strong adherence to both the frame and the prompt.

Examples to try

Each example below ran end to end through this endpoint at the settings shown, animating a supplied frame while keeping its composition, subjects and aspect ratio; only the prompt and frames change between them.

• •Animate a still into complex fluid motion (768P, 6s). Prompt: "Dense white fog billows from the tilted steel flask and cascades over the edge of the stainless bench, rolling low across the floor and swallowing the row of glass beakers as frost blooms up the flask under cold blue light." The cryogenics bench comes alive as liquid-nitrogen fog billows and rolls across the floor while the glass beakers stay sharp and reflective. Keep one primary action per prompt.
• •First-to-last-frame interpolation (480P, 5s). Prompt: "The tight crimson rosebud unfurls petal by petal into a full open bloom, the outer petals curling back first as dew beads tremble on them against the dark background." Supplying both a first frame (tight bud) and a last frame (full bloom) makes the clip bridge the two petal by petal. Give the frames matching framing and subject for a clean path.
• •Fast articulated action, Max tier (480P, 5s). Prompt: "The blue rally car power-slides through the gravel hairpin, its back end stepping out as a spray of stones and dust fans off the rear tyres while a chase camera tracks alongside through the drift." The car throws a dense spray of stones and dust through the drift and holds together, no smearing. Put the moving element already in motion in the first frame; a 480P/5s draft is enough to check the action.

More motion prompts to run over your own frames: "steam curls off a fresh espresso as the crema settles", "a kite lifts off the sand and climbs into the wind", "sparks arc off an angle grinder biting into steel".

Does it take an image?

Yes. Supply a first frame, a last frame, or both.

What happens when I provide both frames?

The clip interpolates between the first and last frame.

Does it generate audio?

Yes, H3 produces native stereo audio alongside the video.

What resolutions and durations are supported?

480P or 768P, from 5 to 15 seconds.

How is it different from H3 Max Text to Video?

This endpoint animates a supplied frame; the text-to-video tier generates a clip from a prompt alone.

When should I use the quality prompt-expansion mode?

For sparse prompts that need enriching; keep balanced for detailed, well-written prompts.