Subject
Name the person, object, animal, or place that carries the shot. Include one or two identifying details only when they need to remain visible.
Learn how to write effective MiniMax H3 prompts with a clear structure, copyable templates, and tested examples for cinematic scenes, camera movement, characters, products, image animation, and reference-guided AI videos.
A useful prompt describes a shot that can be seen and reviewed. Start with the subject and action, then add only the environment, camera, lighting, style, and constraints that change the result.
Subject + Action + Environment + Camera + Lighting + Style + Motion
Write the subject and action first because they define what success looks like. “A runner crosses a bridge” is easier to evaluate than “an epic inspirational world.” The environment should support the action, while the camera instruction explains how the viewer experiences it. Lighting and style shape the image, but they should not replace a clear event.
Finish with constraints when identity, product shape, framing, or continuity matters. For example: “preserve the logo and bottle proportions,” “keep the same face and jacket,” or “no scene cut.” If a sentence does not affect a visible decision, remove it. Clear prompting is less about length and more about giving every phrase one job.
Use these elements as a checklist, not a requirement to make every prompt long. Reference-based modes may already provide the subject, environment, or style.
Name the person, object, animal, or place that carries the shot. Include one or two identifying details only when they need to remain visible.
Use a concrete verb and a readable sequence. Turning, opening, walking, lifting, drifting, or looking gives the model a clearer motion target than an abstract mood.
Describe where the camera starts, how it travels, and where it finishes. Add restrictions such as stable horizon, locked frame, or no sudden zoom when they matter.
Give the subject a physical place with useful depth cues: a wet street, quiet workshop, open coastline, gallery, studio surface, or crowded plaza.
Name the source and quality of light, such as diffused window light, hard noon sun, neon reflections, or a narrow studio sweep.
Choose a visual treatment, then protect non-negotiable identity, shape, branding, composition, timing, or continuity details.
Choose the template that matches your input. Replace every bracketed phrase, keep one main action and one camera path, then remove instructions that do not change a visible part of the result.
[Subject] [performs one visible action] in [environment]. [Camera starting position], then [camera movement] and finish on [final framing]. [Lighting source and quality], [visual style], natural motion, [details that must remain consistent], no scene cut.
Use this when the model must build the complete shot from text. Replace every bracketed phrase with one concrete visual decision.
Animate the subject with [specific movement]. The camera [camera movement or remains locked]. Preserve the face, clothing, composition, colors, and background from the source image. Add only [small environmental motion]. Natural timing, stable geometry, no scene cut.
The image already defines appearance and composition, so the prompt concentrates on motion and preservation.
Use the image reference for subject identity and wardrobe. Use the video reference for movement and timing. Create a new shot in [environment] with [camera direction]. Preserve facial features, body proportions, and clothing details. Do not copy unrelated background elements from the motion reference.
Assign one clear role to each source before describing the new shot and its non-negotiable details.
Open the text-to-video workspace with MiniMax H3 selected, then build one visible shot from the subject, action, camera, and lighting decisions above.
Cinematic prompting works when the camera and action create a shot, not when the prompt simply repeats the word cinematic.
A solitary astronaut crosses a wind-shaped ice field at blue hour, low tracking shot, fine snow moving across the lens, restrained camera shake, cold cyan light with a warm horizon, cinematic realism
One subject action, one camera path, environmental motion, and a controlled lighting contrast.
A woman in a red raincoat walks through a neon-lit Tokyo side street, rain reflecting across the pavement, slow dolly-in, shallow depth of field, passing umbrellas in the foreground, natural walking motion, cinematic night photography
Uses foreground movement and reflections to create depth without adding another scene.
Begin behind a half-open workshop door, push slowly into a quiet room where a violin maker lifts a finished instrument toward the window, warm morning light, floating dust, subtle handheld movement, finish on a medium close-up
Defines a readable beginning, camera transition, subject action, and final composition.
Product prompts should protect geometry, labels, and material while giving light and camera movement enough room to reveal the object.
A brushed steel watch rests on black stone as a narrow band of light travels across the dial, slow 30-degree orbit, keep the logo, hands, and case geometry unchanged, premium studio reflections, clean dark background
Protects brand and geometry while limiting movement to a controlled reveal.
A frosted glass serum bottle stands on pale stone beside a shallow pool, a soft ripple moves through the reflection, gradual push-in, preserve the label and bottle proportions, diffused daylight, clean editorial product video
Separates environmental motion from product details that must remain stable.
When a face, garment, or performance must remain consistent, say what should stay fixed before describing expression and movement.
The same young chef from the reference turns toward camera, smiles naturally, and plates the final dish, preserve facial features and apron details, medium close-up, gentle handheld movement, warm restaurant lighting
States identity constraints before expression, action, framing, and light.
A model in a structured silver jacket walks through a white gallery, fabric moving naturally with each step, preserve the garment shape and face, camera tracks parallel at waist height, soft overhead light, no scene cut
Connects garment consistency to a simple lateral movement the viewer can judge.
A detective enters a dim archive, pauses, then looks toward a moving desk lamp, keep the same face, coat, and hairstyle, slow pan from the doorway to a medium shot, restrained expression, atmospheric dust, realistic motion
Uses a short sequence while keeping identity and camera direction explicit.
A camera instruction is most useful when it has a start, path, subject relationship, and finish. Avoid combining several moves that cannot fit the clip.
Begin on a wide coastal cliff at sunrise, push slowly toward the cyclist, then arc left as the rider passes camera, stable horizon, natural motion blur, no sudden zoom, no scene cut
Defines the start, transition, end, and restrictions in chronological order.
Start directly above a circular market plaza, descend smoothly while rotating clockwise, reveal the central fountain and surrounding stalls, maintain a stable center point, midday sunlight, realistic crowd movement, finish at eye level
Anchors a complex camera move to one fixed visual target so the shot stays legible.
Use a reference image to protect appearance, a video to guide movement, or audio to shape timing, then let the prompt define the new shot.
Translate “dynamic,” “epic,” or “beautiful” into visible action, camera position, lighting, scale, weather, or pacing. Concrete direction is easier to execute and review.
Fit the action to the selected duration. A short clip usually benefits from one main subject movement and one camera path rather than a complete story.
State the opening frame, movement, and final composition. This is clearer than listing dolly, orbit, pan, and zoom as unrelated style words.
Describe where light comes from and how it behaves: window light across a face, a moving reflection on metal, or neon reflected in wet pavement.
Name the face, wardrobe, logo, product proportions, palette, or background element that must not drift, especially when using source media.
On the reference generator, explain whether each source controls identity, motion, composition, style, or timing. Do not ask every source to control everything.
Use short, specific constraints to protect important details. State what should remain stable or which visible behavior should not happen instead of adding a long generic list of defects.
Use constraints such as “preserve facial features, hairstyle, clothing, and body proportions” when a character must remain recognizable.
State “keep the logo, label, text, color, and product geometry unchanged” before adding camera or environmental motion.
Add “stable horizon,” “no sudden zoom,” “no camera shake,” or “no scene cut” only when that behavior would break the intended shot.
Do not paste dozens of unrelated defects into every prompt. Too many restrictions compete with the subject, action, and camera direction.
Prefer “natural hand movement” or “consistent facial features” over a long list of possible anatomy failures. Positive direction gives the model a clearer target.
Text-to-video needs a complete visual brief. Image and reference modes benefit more from preservation rules because the source already defines appearance.
Text, image, and reference workflows need different amounts of description because they start with different information.
Describe the complete shot because the model starts without a source frame. Include subject, setting, motion, camera, lighting, and finish.
Try a text prompt →Do not waste the prompt redescribing the image. Focus on subject movement, camera movement, timing, and the visible details that must remain unchanged.
Animate an image →Assign each image, video, or audio reference a role, then describe the new action and the constraints that connect those sources.
Use references →Start with the subject and visible action. Add the environment, camera path, lighting, style, and the details that must remain stable. Keep the shot focused enough to complete within the selected duration.
A reliable order is subject, action, environment, camera, lighting, visual treatment, motion quality, and constraints. You can shorten the structure when a reference image already defines appearance and composition.
You can describe camera behavior with practical film language such as push-in, pull-back, pan, orbit, tracking shot, overhead descent, or locked camera. Describe the start and finish instead of listing moves without order.
Use enough detail to remove ambiguity, not enough to describe every pixel. One focused paragraph is often easier to execute and review than several competing scenes, styles, or camera moves.
Yes. On image-to-video and reference-to-video pages, use the prompt to describe movement and constraints rather than repeating what is already visible in the source image.
Write visible constraints inside the main prompt, such as no scene cut, no sudden zoom, stable horizon, preserve facial features, or keep the logo unchanged. Use only restrictions that protect an important part of the shot.
Yes. Keep the template structure, replace every bracketed phrase, and remove any clause that does not apply. A template is a checklist for visible decisions, not text that must stay unchanged.
Every example on this page opens the text-to-video generator with the prompt and MiniMax H3 selection prepared. You can edit the wording and review credits before generating.
Open the generator, edit an example, review the credit estimate, and create when the shot is ready.
Move from research to a relevant generator, prompt resource, or pricing page without restarting your workflow.
Create with image, video, and audio references.
Follow the complete generation workflow.
Understand the speed-focused search intent and availability.
Camera movement, shot composition, and motion control prompting.
Transform an existing clip with a video-first reference generator.
Reapply movement from a video reference onto a new subject.
See every online input path in one workflow map.
Understand LoRA availability and the closest online option.
See what a ComfyUI-based path involves and the online alternative.
What API access means here and the online generator alternative.
VRAM, GPU, Mac, AMD, and model size questions in one place.
What running MiniMax H3 locally would involve, mapped out.
Where to look for repositories, workflows, and implementations.
What to look for on Hugging Face: weights, variants, and LoRA.
Open the generator with MiniMax H3 selected.
Compare plans and one-time credit packs.