MiniMax H3 Prompt Generator
Turn a one-line idea into a production-ready MiniMax H3 (Hailuo 3.0) prompt — timecoded shot list, camera direction, native audio and negative constraints. Free, no sign-up for the builder.
Describe your shot
No references. The words carry everything.
H3 generates sound in the same pass — leaving this blank is still a choice.
Your H3 prompt
Describe your scene on the left and the prompt appears here instantly — no sign-up needed. Want AI to expand it into a full shot list? That part needs a free account.
Why H3 prompts look different
MiniMax H3 (Hailuo 3.0) reads text, images, video and audio in one pass and returns 5–15s clips at up to 2K with natively synced sound. It follows very long prompts — working examples in community collections run past 6,000 characters. The prompts that hold up read like production paperwork rather than a caption: a timecoded shot list, an explicit job for every reference, deliberate sound design, and a list of things that must not happen.
t2va, i2va and fl2va need different prompts
t2va (text only) puts the whole burden on words, so camera, lighting and audio all have to be stated. i2va accepts up to nine reference images — the win comes from assigning each image a role instead of describing them as a group. fl2va is given a first and last frame and must invent the motion between, so describe the transition as beats rather than a destination. You use either nine references or a first/last pair, never both.
Write actions as sequences, not labels
"A man jumping" names an activity but not how the body moves. Split important movement into two or three chronological beats and order them with first / then / as / finally. The sequence has to fit the clip length, so a short generation should carry one primary action rather than a whole narrative. If a person or product recurs, restate the identity anchors — face, hair, clothing, proportions, accessories — in every beat.