MiniMax Hailuo H3 Max Reference to Video

MiniMax Hailuo H3 Max builds a video from your reference material. Upload up to 12 images, video clips or audio files and describe the shot — the model keeps the characters, style and motion you supplied while generating up to 15 seconds of new footage.

📄 About MiniMax Hailuo H3 Max Reference to Video

MiniMax Hailuo H3 Max Reference to Video generates footage that stays faithful to material you already have. Instead of describing a character from scratch, you upload it — a photo of a person, a product shot, a style frame, a short clip whose motion you want echoed, even an audio bed to match the pacing. The model reads up to twelve reference files at once and composes a new shot of up to fifteen seconds around them. This matters most when consistency is the job. Marketing teams keep the same spokesperson across a campaign. Game and animation studios hold a character design steady between shots. Product sellers keep the exact packaging, label and colour of an item while placing it in new scenes. Because the references carry the identity, your prompt is free to describe camera work, lighting and action rather than re-describing what things look like. Output runs at 480P or 768P with a choice of seven aspect ratios, including an adaptive mode that follows the shape of your references. Duration is selectable from five to fifteen seconds, and pricing scales with both length and resolution — a five second 480P clip is the cheapest way to test an idea before committing to a longer, sharper render.

✨ Key Features

Up to 12 reference files - images, video clips and audio combined

Keeps character identity, style and product detail across shots

5 to 15 second output, selectable per generation

480P and 768P output with seven aspect ratios

Adaptive aspect ratio follows your reference material

Prompt expansion modes: balanced or cinematic quality

💡 Use Cases

⚡Keeping one spokesperson or model consistent across a campaign

⚡Animating a product while preserving its exact packaging and labels

⚡Holding a character design steady between animation shots

⚡Rebuilding a scene in a new style from a reference frame

⚡Matching the motion of an existing clip with new subject matter

⚡Producing vertical social cuts from landscape reference footage

🎯 Best For

🎯 Teams who need the same character, product or style to survive across multiple generated videos

👍 Pros

✓Reference-driven consistency without training a custom model

✓Accepts images, video and audio references in one request

✓Fifteen second maximum is longer than most reference-to-video models

✓Adaptive aspect ratio removes guesswork about framing

⚠️ Considerations

△Reference video and audio are capped at 15 seconds combined

△Maximum resolution is 768P, so it is not a finishing-quality render

△Large reference images cost extra beyond the free token allowance

📚 How to Use MiniMax Hailuo H3 Max Reference to Video

Upload your reference images - the character, product or style you want kept

Optionally add a reference video whose motion should be echoed, or audio to set the pacing

Describe the shot: camera movement, action and lighting, not what things look like

Pick duration, resolution and aspect ratio, then generate

Ready to try MiniMax Hailuo H3 Max Reference to Video?

Get 10 free credits — no credit card required

Start Free →

Frequently Asked Questions

Twelve in total, counted across images, videos and audio together. Reference videos and audio must each be between 2 and 15 seconds, and their combined length cannot exceed 15 seconds.

That is what the model is built for. Supplying a clear reference image of the character is far more reliable than describing them in the prompt, and it holds up across separate generations so you can build a sequence.

Balanced stays close to the wording you wrote. Quality lets the model add cinematic detail - lighting, camera language, atmosphere - which usually looks richer but drifts further from your exact instructions.

Start at 480P while you are iterating; it costs less per second. Switch to 768P once the composition and motion are right. The price scales with both resolution and duration.

Yes. Uploading a short audio clip lets the model match the pacing and rhythm of the generated motion to your track, which is useful for music-led social cuts.

MiniMax Hailuo H3 Max Reference to Video

More Video Generation Models