Native 16-megapixel output
Generate 4096×4096 squares or 6272×2688 ultrawides directly — no upscaler, no detail loss.
MiniMax
16-megapixel images with Minimax H3 Image
Generate poster-ready visuals directly at 4K, or edit with up to 16 reference images — readable Chinese & English text, ten aspect ratios and six-step sampling.
Minimax H3 Image Studio
text & edit to image
Describe any image — posters with readable text, product shots, puzzles — and H3 Image renders it at up to 16 megapixels.
Click or drop images here
1–16 images (JPG/PNG/WEBP, max 20MB each). Refer to them by order in the prompt, e.g. “apply the style of Picture 2 to the product in Picture 1”.Up to 5,000 characters, Chinese or English. Quote any text that must appear inside the image.
Estimated cost 20 gems
Your image will appear here
What is Minimax H3 Image?
MiniMax’s open-source 33.1B image model — native 16-megapixel generation and multi-reference editing in one checkpoint.
16 MPNative 4K output
16 refsMulti-reference editing
10 ratios1:1 to 21:9
BilingualCN & EN text rendering
Why creators use Minimax H3 Image
Large-format visuals and precise multi-reference edits without an upscaling pass.
16-reference editing
Combine products, characters and style frames in one edit; each reference keeps its upload order.
Text that reads
Posters, diagrams and infographics with legible Chinese and English typography inside the image.
Six-step sampling
A distilled sampler with fixed steps and guidance — fewer knobs, consistent speed.
Prompt rewriting
Short prompts are expanded into detailed English descriptions by default for better results.
Wide campaign assets
21:9 hero banners and 9:16 verticals at the same pixel budget as a square image.
How to create an image
Pick a mode, configure the output and generate.
1. Choose a mode
Text to Image for new visuals, or Edit Image with 1–16 references for edits and composites.
2. Write the prompt
Describe the whole frame — subject, setting, lighting and any text — in Chinese or English.
3. Configure the output
Set the aspect ratio (auto picks one for you) and the size tier: 1K, 2K or 4K.
4. Generate and download
The studio estimates your cost live. Preview the result inline and download the finished image.
Minimax H3 Image FAQ
Answers to common questions about inputs, settings and pricing.
What is Minimax H3 Image?
Minimax H3 Image is MiniMax’s open-source image model. One checkpoint generates and edits images at up to 16 megapixels directly — from 4096×4096 squares to 6272×2688 ultrawide frames — without an upscaler.
What is the difference between Text to Image and Edit Image?
Text to Image generates a new picture from your prompt. Edit Image takes 1 to 16 reference images and follows your instruction — refer to each one by upload order, for example “keep the product from Picture 1 and apply the style of Picture 2”.
What resolutions and aspect ratios are supported?
Three tiers: 1K (about 1 megapixel), 2K (about 4, the default) and 4K (about 16). Ten aspect ratios from 1:1 to 21:9 and 9:16, plus auto, which lets the prompt rewriter choose.
How does billing work?
A flat per-image rate by size tier: 10 gems for 1K, 20 for 2K and 30 for 4K. Reference images in Edit mode are free — you only pay for the output image.
Does it support negative prompts or custom steps?
No. The model is distilled to six sampling steps with no classifier-free guidance; steps and guidance are fixed and negative prompts are not supported.
What is prompt expansion?
On by default — Kie rewrites your prompt into a detailed English description of the whole image, which improves quality. Turn it off for literal instructions; it is also 10–20 seconds faster.
Which languages can I prompt in?
Chinese or English, up to 5,000 characters. The model renders readable text inside images in both languages.
Ready to create your next visual?
Start with a prompt or a set of references and generate 16-megapixel images with Minimax H3 Image.
Start creating with H3 Image