MiniMax H3 source prompts: choose the right input mode
628 source examples; not independently reproduced and no identical-result promise.
How to choose an example
This collection contains the main H3 model, not Fast or Max. Its current Studio settings are 4–15 seconds and 768P/2K.
Choose text, frame or reference from your actual inputs; frame roles and reference roles belong to separate tasks.
Check each model, input role and known setting. Attribution is not a media reuse grant; supply your own inputs.
- Pink-and-Cyan Afterimages on a Lemon Stage
A red-haired anime dancer performs fast choreography on a pure yellow stage as pink and cyan afterimages peel from her limbs and snap back together. Native-video output: 15s · 2560
MiniMax H3 · reference-to-video · 0 recorded input roles
- A Last Handhold on a Blizzard Cliff
A climber slips on a snowy cliff, drops until her rope catches, and finally secures the ledge below a buried mountain entrance. Native-video output: 15s · 3840×2160. The creator pu
MiniMax H3 · reference-to-video · 0 recorded input roles
- A Teasing Conversation by the Starship Window
Two women and a cat share a warm spacecraft cabin as the camera moves between a mug, subtle expressions, and small gestures during a teasing exchange. Native-video output: 15s · 38
MiniMax H3 · reference-to-video · 0 recorded input roles
- A City Painted into Existence
Oil-paint ridges turn into streets and landmarks before resolving into an Edinburgh poster. The published prompt includes city and quote variables.
MiniMax H3 · text-to-video · 0 recorded input roles
- Building a Character Bust from Scratch
Hands assemble an armature, sculpt a face and add paint, hair and accessories to a finished character bust. The prompt requires a character reference image.
MiniMax H3 · text-to-video · 0 recorded input roles
- Glass Shards Reassemble into SHATTER
A glass monolith fractures into flying shards that assemble into a readable title, followed by another burst of fragments.
MiniMax H3 · text-to-video · 0 recorded input roles
- Escaping a Manga Book, Meeting Another Self
A manga character emerges from the pages into live action and discovers another version still inside. The prompt references multiple images; some panel and speech-bubble exclusions
MiniMax H3 · reference-to-video · 5 recorded input roles
- A Live-Action Muse Meets a Tiny 2D Fairy
A tiny drawn fairy interacts with a live-action woman through palm contact and mirrored poses. Character reference images are required by the published prompt.
MiniMax H3 · text-to-video · 0 recorded input roles
- Jade Panic: A Yellow-and-Green Character Entrance
A stylized anime character moves through bold yellow-and-green graphic compositions, action poses and impact typography. The prompt requires a character reference.
MiniMax H3 · text-to-video · 0 recorded input roles
- A Coastal Walk in Knitwear
A coastal walk alternates mountain and sea views with close portraits.
MiniMax H3 · text-to-video · 0 recorded input roles
- From Hagia Sophia to the Bosphorus
A traveler moves from city landmarks to a ferry before a sunset skyline closes the sequence.
MiniMax H3 · text-to-video · 0 recorded input roles
- Pink Seoul Fashion in Graphic Cuts
Pink streetwear and oversized type alternate as the model moves between city scenes and solid backdrops.
MiniMax H3 · text-to-video · 0 recorded input roles