Reference-driven scenes
Builds a new scene around your references instead of animating a single still, so the setting, lighting and camera are all generated fresh.
AI Zombie Video2 photos · 15s · from 121 credits
Burning Bridges AI1 photo · 15s · from 121 credits
Wipe Me Down Dance AI1 photo · 13s · from 105 credits
AI Vampire Transformation1 photo · 5s · from 21 credits
La Tormenta de Arena AI2 photos · 15s · from 121 credits
AI Witch Transformation1 photo · 5s · from 21 creditsMiniMax H3 (the latest Hailuo model) is built around reference images. Give it photos of a person, a pet, a product or a location, describe what should happen, and it generates a scene that keeps those references recognisable — with sound, at 768p or 2K.
References in, video out
Add 1 to 9 images: the people, pets, products or places that should appear in the video.
Say what happens, where and how the camera moves, for example "the woman walks along a beach at sunset and waves".
768p for social clips or 2K for sharper delivery, and any duration from 4 to 15 seconds.
H3 always returns a clip with sound. Download the MP4 when it is ready.
Builds a new scene around your references instead of animating a single still, so the setting, lighting and camera are all generated fresh.
Walking, waving, turning and hair in the wind look smooth and believable, even in wide shots.
Every H3 clip comes with generated sound — ambience and effects matched to the scene.
Choose 768p for quick social posts or 2K when the clip needs to hold up on a big screen.
20 credits per second at 768p, 33 at 2K, plus 10 credits per reference image.
H3 is one of the models behind Viberot's viral video templates, such as Hotel Lobby AI.
Credits for a failed generation are returned automatically.
20 credits per second at 768p, 33 at 2K, plus 10 credits per reference image. The price is shown before you generate, and failed generations are refunded.
Up to 9 reference images, 768p or 2K, sound included.