MiniMax H3
Prepare text, starting and ending frames, or a separate reference-video brief without mixing incompatible inputs. Describe one action and a camera direction for a focused video brief.
Video Generator
Plan the shot before generating
Start with the subject and the movement that matters most. Keep reference images consistent with your intended framing. Review the selected duration, resolution and credit quote before each video submission.
Can I start with welcome credits?
Use the 3-credit welcome offer for one GPT Image 2 image at 1K, then decide whether to turn that frame into a video. Video generation needs sufficient paid credits. Your plan, balance and private history belong to Grok Video.
See video creation plansMiniMax H3 visual examples
Explore the examples, compare the visual choices, and open a card for a prompt or input guide. Published source records and editorial reference prompts are distinguished below. File dimensions and duration describe the downloaded asset.
Prepare the input
Start with a scene brief. For editing or animation, add the matching reference inputs requested by the generator.
What to compare
Play the clip and inspect the beginning, middle and end. Compare the subject, motion, camera path and sound; a thumbnail cannot establish temporal consistency.
Example boundaries
These are selected external examples, not a promise that a prompt will reproduce the same result. Model versions, references and settings can change the outcome.
Source review: 2026-09-08. Preview dimensions are measured from the original files.
Input reference
Hooded character running through a street
For framing or visual direction; not a verified output from this tool.
2560 × 1440
Prompt and settings
Editorial reference prompt
Create a composition showing hooded character running through a street. Specify the main subject, its placement and the lighting. Keep the background coherent and check small details at the intended output size.
Written from the visible scene as a starting point; this is not an original generation log.
Model example
Man holding a small dog
MiniMax H3 model-page example.
2560 × 1440 · 6.6 s
Prompt and settings
Editorial reference prompt
Create a short video showing man holding a small dog. Establish the subject before introducing movement. Use one readable camera move and keep the scene consistent through the final frame.
The source pairs this asset with a published prompt or playground settings. The editable brief here is an editorial adaptation; open the source for its complete record.
Input reference
Armored characters facing each other
For framing or visual direction; not a verified output from this tool.
2560 × 1440
Prompt and settings
Editorial reference prompt
Create a composition showing armored characters facing each other. Specify the main subject, its placement and the lighting. Keep the background coherent and check small details at the intended output size.
Written from the visible scene as a starting point; this is not an original generation log.
Input reference
Athlete seated in a stadium
For framing or visual direction; not a verified output from this tool.
1456 × 816
Prompt and settings
Editorial reference prompt
Create a composition showing athlete seated in a stadium. Specify the main subject, its placement and the lighting. Keep the background coherent and check small details at the intended output size.
Written from the visible scene as a starting point; this is not an original generation log.
Model example
Hand-drawn kitchen comedy scene
MiniMax H3 model-page example.
2560 × 1440 · 15.1 s
Open original videoPrompt and settings
Editorial reference prompt
Create a short video showing hand-drawn kitchen comedy scene. Establish the subject before introducing movement. Use one readable camera move and keep the scene consistent through the final frame.
Written from the visible scene as a starting point; this is not an original generation log.
Model example
Conversation beside a kitchen sink
MiniMax H3 model-page example.
2560 × 1440 · 15.1 s
Prompt and settings
Editorial reference prompt
Create a short video showing conversation beside a kitchen sink. Establish the subject before introducing movement. Use one readable camera move and keep the scene consistent through the final frame.
The source pairs this asset with a published prompt or playground settings. The editable brief here is an editorial adaptation; open the source for its complete record.
Model example
Crafted paper village in soft light
MiniMax H3 model-page example.
1920 × 1440 · 15.1 s
Prompt and settings
Editorial reference prompt
Create a short video showing crafted paper village in soft light. Establish the subject before introducing movement. Use one readable camera move and keep the scene consistent through the final frame.
Written from the visible scene as a starting point; this is not an original generation log.
Input reference
Two people in a kitchen
For framing or visual direction; not a verified output from this tool.
2560 × 1440
Prompt and settings
Editorial reference prompt
Create a composition showing two people in a kitchen. Specify the main subject, its placement and the lighting. Keep the background coherent and check small details at the intended output size.
Written from the visible scene as a starting point; this is not an original generation log.
Model example
Toy character seated in a library
MiniMax H3 model-page example.
1920 × 1440 · 6.6 s
Prompt and settings
Editorial reference prompt
Create a short video showing toy character seated in a library. Establish the subject before introducing movement. Use one readable camera move and keep the scene consistent through the final frame.
Written from the visible scene as a starting point; this is not an original generation log.
What can you create with MiniMax H3?
Create a short scene with MiniMax H3 using text, frame inputs or a separate reference-video workflow. Pick the input mode that matches your material before planning the shot. On Grok Video, start with an opening shot and a single camera movement. Keep your source material and intended result together in the generator above.
Prepare your inputs
In text-to-video, describe the subject, action, camera and any intended sound. For image-to-video, supply the frame inputs requested by the form. For reference-to-video, use a short source clip and explain the visual direction it should guide.
A product reveal needs a readable silhouette and a deliberate camera path. A dialogue scene needs clear speakers and short lines. A reference-led shot works best when the subject and movement in the source are compatible with the new brief.
From brief to result
- Select the mode and describe your intended result in the generator on this page.
- Add references you own or have permission to use, then choose output settings.
- Review the exact credit quote before submitting. Availability and balance checks happen before a generation request.
- When a task completes, review the output before downloading it or starting another variation. A pending task is not a completed result.
Limits to check before generating
Frame inputs and reference-video inputs belong to different modes. Keep them separate. Available output settings and the quote depend on that choice. Check facial details, scene transitions and any dialogue or music in the finished video.
Do not submit confidential material without the necessary permission. An accepted upload does not guarantee provider approval or a usable output. If a request fails, keep the task record and check its credit refund rather than repeatedly submitting the same input.
Credits and availability
The selected duration, resolution and references determine the quote. A reference video can affect how the request is priced. Video creation is not covered by the welcome still-image offer; inspect the current estimate before submitting.
A model marked unavailable cannot start a paid task. It is not a reason to buy credits until generation has been enabled. Plans, balances, and private history apply to Grok Video.
Can I change models without starting over?
Keep the brief, then review the new model's settings. Different models accept different references and durations, so check the required inputs and quote again before submitting.
Provider contract reviewed September 8, 2026. Official MiniMax H3 API documentation. Upload limits on this website may be narrower than the provider's maximum.
Explore image and video models
- GPT Image 2Image
- Grok Imagine Image 2.0Image
- Nano Banana 2Image
- Grok Imagine Video 1.5 PreviewVideo
- MiniMax H3Video
- Veo 3.1Video
- Kling O3Video
- Gemini Omni 1.1 FlashVideo
- Seedance 2.5Video
- Seedream 5.0 ProImage



