MiniMax Video Generator — Create Clips with MiniMax H3
MiniMax H3 is the video model from MiniMax: describe a scene or upload photos and get a short clip back. On Tokion it runs on the same pay-as-you-go balance as chat, images and the API.
- $0.16from per video
- 1Versions and prices
Versions and prices
| Version | Best for | Price |
|---|---|---|
| MiniMax H3 Accepts a reference photo | Текст 1–15 с, фото 1–10 с · 480p/768p/1080p | $0.16 per video |
Made with MiniMax
MiniMax video generator vs. MiniMax chat models
MiniMax builds both language models and a video model, and it's easy to mix them up. This page is about video only: MiniMax H3 turns prompts and images into clips. If you're after the MiniMax chat models for writing, coding or document work, you'll find them in the Tokion model catalog at /models, with their own per-token pricing.
Both live on the same Tokion balance, so you can use a MiniMax chat model to draft a video concept and then send that concept to H3 without topping up anything separately.
What MiniMax H3 can do
H3 is a single model with two ways in. Text to video: write a scene and get a clip from 1 to 15 seconds. Image to video: attach up to nine images and the model builds the clip from them, with a length from 1 to 10 seconds.
You can render in 480p, 768p or 1080p. 480p is the default and the most budget-friendly setting for drafts. One rule to remember: 1080p is only available when you attach at least one image. Prompt-only clips top out at 768p.
That spread of options suits an iterate-then-finalize workflow. Burn through a few cheap 480p takes to find the shot, then re-run the keeper at a higher resolution.
How to use MiniMax H3 on Tokion
Open the Video section in the Tokion web app or the Telegram bot and choose MiniMax H3 in the model selector. Enter your prompt, attach images if you want, set duration and resolution, and start. The clip arrives in the chat, ready to download or share.
Billing is per clip with no subscription, and the cost scales with duration and resolution; the version table on this page lists current rates. Your one Tokion balance covers chat with dozens of language models, image tools, video models and the API.
Developers can call MiniMax H3 through the Tokion API with the same key used for every other model. Head to /api-landing for the details.
Getting better results from MiniMax AI video
Write each prompt like a single camera shot. Name the subject, one action, the lighting and how the camera moves: "a ceramic mug on a windowsill, steam rising, soft morning light, slow push-in." Short and specific beats long and poetic.
Stick to one action per clip. A five-second video can't carry a whole storyline, and trying to cram one in tends to produce rushed motion. Build longer stories from several short clips in your editor.
For image to video, pick sharp, well-lit photos with a clear subject that isn't cropped at the edges. Whatever is in the image, including noise and blur, shows up in the motion.
Who it's for
Social creators
Fast 480p drafts let you try several ideas for Reels, TikTok or Shorts before you commit. Keep the best take and re-render it at a higher resolution.
Online sellers
Turn product photos from your Amazon, Etsy or Shopify listing into a short motion clip. Attach a clean image to unlock 1080p for the final version.
Small brands and agencies
Animate a hero image into a looping website background or a quick teaser for a launch. Share a moving concept with a client instead of a static mockup.
Developers
One Tokion API key reaches MiniMax H3 and the chat models. Build a bot that writes the caption and returns a matching clip in one flow.
Pro tips
- Draft in 480p first; it's the quickest way to check the model understood your scene.
- Want 1080p? Attach an image; the option only appears in image-to-video mode.
- Need 11 to 15 seconds? Use text to video, since image-based clips cap at 10 seconds.
- Describe camera motion in plain terms: push in, pull out, orbit, tracking shot, static camera.
- Keep prompts to a few focused sentences with one action and clear lighting.
- Change one variable between takes so you know what made the difference.
Good to know
- Image-to-video clips max out at 10 seconds, and text-to-video clips at 15 seconds.
- 1080p requires an attached image; prompt-only clips are limited to 480p or 768p.
- On-screen text and small logos can warp in motion, so add titles in your editor.
- This is a video model, not a chatbot. For text tasks, use the MiniMax chat models in the /models catalog.
Questions
Is MiniMax H3 the same as the MiniMax chat models?
No. Same developer, different models. H3 makes video; the MiniMax language models handle text and are listed in the /models catalog. Both run on your one Tokion balance.
How long can a MiniMax clip be?
1 to 15 seconds from a text prompt, or 1 to 10 seconds when you attach images.
Why can't I select 1080p?
1080p is only offered in image-to-video mode. Attach an image and the option becomes available.
How many images can I attach?
Up to nine per clip. Only upload images you own or have permission to use.
Do I need a subscription?
No. Tokion is pay as you go: each clip is charged to your balance, which also covers chat, image tools and the API.
Can I use MiniMax H3 through an API?
Yes, with the same Tokion API key you use for other models. See /api-landing for the request format.
What if the clip doesn't match my prompt?
Simplify to one action, add a clear camera move and run another 480p take. Every render comes out a little different, so a second try often lands.
Other image and video models
MiniMax Video Generator — Create Clips with MiniMax H3
Try the MiniMax video generator H3: text to video up to 15 seconds or image to video up to 10, in 480p, 768p or 1080p. Pay per clip on one Tokion balance.
Try MiniMax