To use MiniMax H3, choose a generation mode, add your prompt or reference media, configure the real video settings, review the credit estimate, and start the generation. The workflow below takes you from an empty brief to a downloadable result.
Make the input-mode decision first. A text prompt, one source image, and a set of references require different preparation even when the final goal is similar.
Choose text to video when you only have an idea, image to video when the opening composition already exists, or reference to video when identity, movement, composition, or timing needs stronger control. This first decision determines how much visual information the prompt must provide.
Describe the subject and visible action first. Add the environment, camera path, lighting, and the details that must not change. Keep the shot focused enough for the selected duration. If you need a structure or working examples, open the MiniMax H3 Prompt Guide and send an example directly to the generator.
For image to video, upload the still frame that should define the composition. For reference to video, add only sources with a clear job: an image for identity, a video for motion, or audio for timing. Public media sources must use HTTPS URLs; supported images can also be uploaded after sign-in.
MiniMax H3 currently exposes duration, 768P or 2K resolution, and supported aspect ratios. Use a shorter 768P pass while testing a prompt or reference setup. Choose 2K or a longer duration after the direction works. The interface calculates credits from the selected settings before submission.
Review the prompt, references, and visible credit estimate, then generate. The job runs as a tracked task rather than an instant browser effect. Keep the page open while it checks progress. When the result is ready, play it, download it, or open your creation history before deciding whether another pass is necessary.
Choose the mode from the material you already have and the kind of control the project needs.
| Goal | Mode | Start |
|---|---|---|
| Start from a written idea | MiniMax H3 Text to Video | Write a complete shot |
| Animate an approved still | MiniMax H3 Image to Video | Upload one clear image |
| Preserve several references | MiniMax H3 Reference to Video | Assign source roles |
| Research faster generation | MiniMax H3 Turbo | Turbo is not currently selectable |
Start from words, animate an approved frame, or combine references. Pick the input you already trust instead of forcing every job through the same mode.
Good generation setup removes ambiguity before credits are spent.
Write a self-contained shot: subject, visible action, environment, camera, lighting, and constraints. Avoid several unrelated scenes in one short prompt. The model has no source frame, so the prompt carries the complete visual brief.
Use the prompt guide →Choose a sharp image with a readable subject and useful space for movement. The source already defines appearance and composition, so the prompt should focus on motion, camera behavior, timing, and details that must remain unchanged.
Open image to video →Use the smallest useful set of sources. State which image controls identity, which video controls movement, and whether audio influences timing. Competing references make the direction less clear.
Open reference generator →Only use settings shown in the live generator. MiniMax H3 controls currently expose durations from 4 to 15 seconds, 768P and 2K resolution, plus supported landscape, portrait, square, and extended ratios.
Choose enough time for one readable action and camera move. Short passes are useful for prompt tests. Longer clips cost more and work better after the shot direction is stable.
Use 768P while checking identity, composition, or motion. Select 2K when the creative direction is approved and the extra output detail supports editing or delivery.
Match the destination before generating: landscape for widescreen placement, portrait for vertical social video, or square where the product layout requires it. Avoid depending on a later crop.
Define one visible outcome before adding style. A product reveal, character turn, camera orbit, landscape movement, or short social beat gives the generation a clear target. Describe motion in chronological order: opening frame, subject action, camera path, and finishing composition. If the subject must remain stable, name the exact face, garment, logo, product shape, or palette that cannot drift.
Input quality matters. Use sharp source images without heavy compression, tiny subjects, or distracting overlays. For references, assign one job to each source instead of asking every file to control identity, motion, style, and timing at once. After a result, compare subject accuracy, motion, camera, composition, and timing against the same checklist. Change one variable per retry so the next output teaches you what improved.
The MiniMax H3 Prompt Guide includes copyable examples for cinematic, product, character, and camera-led shots. Use those examples as structures, then replace the subject and visible decisions with your own brief.
Use a focused action, a supported aspect ratio, and a lower-cost first pass. Review the result, then change one prompt clause or setting at a time.
MiniMax H3 generations use credits. The live generator calculates the estimate from the current model, duration, resolution, and relevant settings before you submit. Use that visible estimate as the source of truth rather than a fixed number in a tutorial.
For early tests, choose a shorter duration and 768P. When the shot works, decide whether a longer or 2K pass is worth the additional credits. Compare recurring plans and one-time packs before buying capacity.
View MiniMax H3 pricing →Open text to video, keep MiniMax H3 selected, write one clear shot, choose a short duration and 768P for the first test, then review the credit estimate before generating.
Yes. The generation controls, source upload, task status, and finished result are available in the browser. You need an account and enough credits to submit a paid generation.
Use text to video for a new idea, image to video to animate an approved frame, and reference to video when several sources need to guide identity, motion, composition, or timing.
The live MiniMax H3 controls support duration choices from 4 to 15 seconds, 768P or 2K resolution, and the aspect ratios shown in the relevant generator. Available controls are the source of truth.
Use a readable input, write one main action, describe the camera in order, protect important identity or product details, and change one variable between retries.
Generations use credits. Cost depends on the live settings, so check the estimate shown before submission and use the pricing page to compare subscriptions with one-time packs.
Choose a mode, prepare one clear shot, and review the credit estimate before generating.
Move from research to a relevant generator, prompt resource, or pricing page without restarting your workflow.
Create with image, video, and audio references.
Write and test prompts for specific video goals.
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.
