How Do I Prompt MiniMax H3 for Cinematic Video?

Overview

Cinematic is a craft vocabulary, not an adjective. Instead of asking for a “cinematic” clip and hoping, describe the specific choices a cinematographer would make: the shot size, the lens feel, the lighting, and the movement. MiniMax H3 responds better to that concrete language, and so will your edit later when the pieces have to cut together.

Source note: Verified 2026-08-06 against the MiniMax official H3 blog, MiniMax Video Generation API docs, and Hugging Face MiniMax-H3 model page.

Build the prompt in film terms. Name the shot — wide establishing, medium, or close-up — and the camera move, such as a slow dolly-in or a locked-off frame. Describe lighting with intent: soft key light, hard shadows, golden-hour warmth, or moody low-key. Add depth cues like a shallow focus or foreground elements, and keep the motion purposeful rather than frantic. One clear mood per clip reads as cinematic; a pile of competing styles crammed into a few seconds reads as noise no matter how good the model is.

Iterate like a director rather than a slot machine. Generate, watch for the one thing that breaks the illusion, and change only that in the next pass so you learn what each adjustment does. A shot that nails composition and lighting usually beats one crammed with effects.

Exact controls vary by version, so confirm which camera and style parameters your route actually supports, and lean on precise prompt language for everything the parameters do not cover.