Reference-to-video prompts: assign appearance, motion and audio
772 source examples; not independently reproduced and no identical-result promise.
How to choose an example
Assign each owned input a job: images for appearance, video for motion and audio for sound. Avoid asking one source to define unrelated roles.
Compare model and listed inputs before opening a case. Missing source media is disclosed; it is not automatically imported or replaced.
Check each model, input role and known setting. Attribution is not a media reuse grant; supply your own inputs.
- Style Presets
A ten-second trap music video pairs two detective rappers with gritty print textures, bold animated typography, and close-up cuts synchronized to the beat.
MiniMax H3 · reference-to-video · 3 recorded input roles
- Three-reference singing performance
Combine a camera-motion video, a character image and a vocal reference in one singing performance.
MiniMax H3 · reference-to-video · 3 recorded input roles
- Follow the Wind
References a source video, its original background music, and an additional voice timbre to add English dialogue while preserving character identity, wardrobe, framing, and setting
MiniMax H3 · reference-to-video · 3 recorded input roles
- Cloud-Head Subway Transformation
A locked-off, symmetrical subway shot glitches the passengers' heads between human faces and distinct weather clouds, then converges on the reference composition at eight seconds.
MiniMax H3 · reference-to-video · 0 recorded input roles
- Twenty-Second H3 Test
Runs a twenty-second H3 test with the same underlying idea adapted to the H3 guide, providing a traceable comparison with an earlier LTX 2.3 version.
MiniMax H3 · reference-to-video · 1 recorded input roles
- Neon Mountain Hover-Bike Race
Generates a fifteen-second two-rider hover-bike race from a supplied first frame, with text constraints that lock the two riders and two bikes while maintaining high-speed continui
MiniMax H3 · reference-to-video · 0 recorded input roles
- Widescreen Character Introduction
A 15-second widescreen character-introduction clip explicitly tagged MiniMax H3 and Hailuo AI. The creator published the complete prompt in a reply, preserved here verbatim.
MiniMax H3 · reference-to-video · 3 recorded input roles
- A Lacquer-Thread Vase from Clay Rolling to Display
An artisan rolls material into fine strands, applies and carves gold ornament across a red vase, and finally places the completed piece on a display stand. Output: 15s · 2560×1440.
MiniMax H3 · reference-to-video · 3 recorded input roles
- A Green Locomotive Mechanically Transforms into a Robot
Green-and-yellow locomotive 8173 unfolds and reorganizes on a mountain railway, ending as a giant robot that retains the engine's front and livery. Output: 15s · 1536×672. The crea
MiniMax H3 · reference-to-video · 2 recorded input roles
- Two-Part Bedroom Dance
A blue-haired sailor-uniform character performs two connected dance segments in a fixed bedroom shot, with natural hair and skirt motion. Native-video output: 12s · 608×1088. The c
MiniMax H3 · reference-to-video · 0 recorded input roles
- Three Outfit Changes in a Bedroom
A woman opens a garment bag in her bedroom, switches through three outfits, and ends posed to camera. Native-video output: 16s · 1088×720. The creator published the complete prompt
MiniMax H3 · reference-to-video · 0 recorded input roles
- A Natural Walk Through a Pedestrian Street
A referenced woman walks naturally through the specified pedestrian street while the camera follows steadily. Native-video output: 7s · 832×1120. The creator published the complete
MiniMax H3 · reference-to-video · 0 recorded input roles