MiniMax H3, image input, 768p, 2k
MiniMax H3 turns your images into crisp 768p videos with rich, cinematic motion.
A tradução deste card ainda não está disponível; parte do conteúdo aparece em inglês.
MiniMax H3 is a video generation model in the Hailuo family built for image-conditioned outputs. You provide a single image as a starting frame and the model produces a short clip with natural motion, coherent physics, and detailed textures at 768p resolution, with options to upscale to 2K for sharper results.
Use H3 when you already have a visual concept — a character pose, a product shot, a scene mood board — and want to bring it to life as a video. It shines at subtle, cinematic camera moves, atmospheric effects, and stylized environments where consistent style from frame to frame matters more than long durations.
What sets H3 apart is its blend of high base resolution, image-grounded consistency, and clean motion that respects the source composition. Compared to text-only video models, H3 gives creators more directorial control because every clip starts from a visual anchor they've already chosen.
- Image-to-video animation
- Social media short clips
- Cinematic camera moves
- Product reveal videos
- Concept art to motion
Generates clean 768p footage with an option to upscale to 2K for production-ready results.
Faithfully preserves the style, subject, and composition of the source image throughout the clip.
Produces smooth, natural movement with refined lighting and atmospheric depth.
- Choose a source image with a clear subject and uncluttered background for the cleanest motion.
- Specify the exact camera move (push-in, pan, orbit, dolly) in your prompt for predictable results.
- Describe lighting and atmosphere (golden hour, neon rain, misty forest) to elevate cinematic quality.
- Generate several variations and pick the best take — motion outcomes vary between seeds.
- Clips are short, so very long narrative sequences require stitching multiple generations.
- Fine text, small faces, and intricate hand movements can occasionally drift or warp.
- Busy or cluttered source images may reduce motion clarity and coherence.
A massive dragon gliding over a neon-lit Tokyo skyline at night, cinematic tracking shot, volumetric clouds, rain-soaked streets glowing below.
A floating island with waterfalls cascading into the clouds, slow orbit camera, dramatic sunset, birds soaring in the distance.
A lone astronaut walking down a glowing futuristic corridor, dolly shot following from behind, particles drifting in the air, blue and orange neon lighting.
A number that locks the randomness so you can reproduce the same result or tweak variations.
How long the generated clip plays, in seconds.
Toggle to enhance the final output from 768p to 2K resolution.
Shape of the video frame, like 16:9 for widescreen or 9:16 for vertical phone clips.
The type of camera move, such as static, pan, push-in, orbit, or dolly.
How much movement the model adds — low for subtle motion, high for dramatic motion.
Os preços exibidos são o que você paga na Muitto. Sem surpresas.
What input does MiniMax H3 need?
A single image as the starting frame, plus a text prompt describing the desired motion and scene.
What resolution does it output?
Clips are generated at 768p with an option to upscale to 2K.
How long is each generated video?
Each generation produces a short clip of a few seconds, designed for punchy, high-impact moments.