Notice: GPT Image 2.6 Flare and Sunburst are newly launched with a limited-time 20% off, try it now
HiAPI
Blog

Guide

MiniMax H3 Image Guide: Text-to-Image, Multi-Reference Editing, Parameters and Pricing

Which model ID to use, how to set parameters, and what it costs

Nov 12, 2026 3 min read
A dark glass prism refracting magenta light in a dim studio, MiniMax H3 Image guide cover
Model
minimax-h3-image
On this page

MiniMax H3 Image is MiniMax's image model. On HiAPI it has two model IDs: minimax-h3-image/text-to-image (generate from text) and minimax-h3-image/image-to-image (generate or edit with 1–16 reference images). This guide covers which ID to use, the parameters, pricing, and a few things to watch when you call it.

Which ID to use

  • Text only: use text-to-image. The only required field is prompt (up to 5,000 characters).
  • You have images to work from: use image-to-image and also pass image_urls with 1–16 images. Put a product into a new scene, repaint a photo as an illustration or oil painting, or pass several references at once.
  • Both IDs cost the same per output image; image-to-image adds a fee per reference image (see Pricing).

Parameters

  • resolution: 1K / 2K / 4K, default 2K. For example, 2K 1:1 outputs 2048×2048, 1K 9:16 outputs 768×1360, and 4K 16:9 outputs 5472×3072.
  • aspect_ratio: auto, 1:1, 5:4, 4:5, 4:3, 3:4, 3:2, 2:3, 16:9, 9:16, 21:9. Text-to-image defaults to 1:1. Image-to-image defaults to auto, where the model picks a ratio from the prompt and reference images; it does not necessarily match the first reference, so set a ratio if you need a fixed frame.
  • output_format: jpeg (default) or png. PNG files are much larger (a 4K PNG can reach 30 MB), so stick with jpeg unless you need PNG.
  • seed: optional, 0–2147483647. Fix it to reproduce an image.
  • Each request returns one image. Custom width/height, transparent backgrounds, and mask-based inpainting are not supported.

Example request

Create a task:

curl https://api.hiapi.ai/v1/tasks \
  -H "Authorization: Bearer $HIAPI_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "minimax-h3-image/image-to-image",
    "input": {
      "prompt": "Place the serum bottle on a pale stone pedestal, soft side light, clean e-commerce product shot",
      "image_urls": ["https://example.com/bottle.png"],
      "resolution": "2K",
      "aspect_ratio": "3:4"
    }
  }'

The response contains data.taskId. Poll GET https://api.hiapi.ai/v1/tasks/<taskId> until status is success; the image URL is at data.output[0].url. A 2K image typically takes 30–40 seconds. Result URLs expire (see expireAt), so download and store images you want to keep.

Pricing

  • 1K and 2K: $0.0129 per image.
  • 4K: $0.0393 per image.
  • Image-to-image adds $0.0036 per reference image. Example: 2K with 3 references = 0.0129 + 3 × 0.0036 = $0.0237.
  • You are only charged for successful images; failed tasks are fully refunded.

Prompt tips

  • Spell out subject, setting, lighting, style, and composition, in that order.
  • For image-to-image, say what to keep and what to change, e.g. "keep the face and pose, replace the background with warm grey blocks, repaint in oil brushstrokes".
  • Explore compositions at 1K or 2K and render the final at 4K, which costs about 3× as much as 2K.

Related articles

View all articles →

Start building with persistent AI artifacts

Get your API key and call image, video and audio models through one task-based API.

Read the API docs