MiniMax Video Prompt Library

This library brings together community videos made with MiniMax H3 and their prompt details. Browse the results to find a useful starting point for your own scene. Prompt access follows the access level shown on each work. Promptsref is an independent platform and is not affiliated with MiniMax.

Open a video you like and review its prompt and any available generation details. Copy the prompt when access is available, adapt it to your subject, and select MiniMax H3 in the video generator. If the example depends on reference media, supply suitable references of your own. A prompt alone does not reproduce someone else's input files or guarantee the same result.

Describe what viewers should see and hear: the subject, setting, visible action, camera movement, and sound. Write events in the order they happen and finish with the intended ending. Start with one manageable scene, then add details that resolve ambiguity. The generator accepts a text prompt; you do not need to fill out separate JSON fields.

Try: “[Subject] in [setting] performs [action]. The camera [movement and framing]. [Lighting or visual style]. Sound: [dialogue, ambience, or effects]. End with [final pose or composition]. Keep [important details] consistent.” For example: “A ceramic artist lifts a blue bowl from a workbench in a bright studio. The camera slowly moves closer as she turns the bowl toward the window. Soft daylight reveals the glaze. Sound: quiet room ambience and the bowl brushing against her apron, with no music. End with the bowl held still at chest height.” This is a writing example, not a claimed generated result.

For text-to-video, describe the scene and its movement. With a first-frame image, focus on what happens after that image. With first and last frames, explain the transition between them. In reference mode, identify what each image, video, or audio file should contribute. On Promptsref, first/last frames cannot be combined with reference media in the same MiniMax H3 request.

Separate the subject's action from the camera's movement. For example, describe a person walking while the camera tracks beside them. Specify the direction and speed when they matter. For speech, identify the speaker and exact words; describe ambience, action sounds, and background music separately so their roles are clear. Keep the requested speech and action within the selected clip length.

Differences in reference media, settings, and generation variation can change the result. Check those first, then look for competing camera directions, incompatible references, or too many actions. Give each reference one clear purpose and state which character or product details should stay consistent. Change one major instruction at a time so you can see what helped.

The current MiniMax H3 generator on Promptsref supports clips from 4 to 15 seconds at 768P or 2K. Prompts can contain up to 7,000 characters. Select duration and resolution in the generator; writing a different value in the prompt does not change those settings.

Generating a MiniMax H3 video uses credits. The cost depends on the selected duration and resolution, and can also change with reference video duration and additional images. Review the displayed credit cost before generating. Prompt access in the library and video generation are separate parts of the workflow.