MiniMax H3 Image
Neural Network
MiniMax H3 Image is a MiniMax model for generating and editing images from text and references.
Specifications
- Input
- text, images
- Output
- images
- Resolution
- 1K, 2K, 4K
- Aspect ratio
- 1:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 16:9, 9:16, 21:9
import OpenAI from 'openai';
const openai = new OpenAI({
apiKey: '<your bothub access token>',
baseURL: 'https://openai.bothub.chat/v1'
});
async function main() {
const response = await openai.images.generate({
model: 'minimax-h3-image',
prompt: 'a white siamese cat',
n: 1,
size: '1024x1024'
});
const imageUrl = response.data[0].url;
}
main();
How it works MiniMax H3 Image?
MiniMax H3 Image is a MiniMax model that covers both sides of image work: it generates images from text descriptions and edits or assembles compositions from existing materials. You can input 1 to 16 references, which is convenient for maintaining a series style, transferring a solution from a sample, or combining multiple subjects into one frame. One tool handles both creation from scratch and refinement. Through BotHub, you work with the model without a VPN or foreign card, pay in rubles, and only for what you use: no subscription is needed, and tokens do not expire. MiniMax H3 Image opens in the same window as other models in the catalog, making it easy to compare results with alternatives and switch without changing the integration. For teams, contracts, invoices, electronic document management, and limits are available; for development, a unified API is provided, and data is encrypted and not saved. Scenarios: illustrations and covers for articles and social media, advertising layouts based on brand books using references, assembling compositions from multiple sources, refining finished images, and creating series of frames in a unified style.Frequently asked questions about MiniMax H3 Image
You can use generated results for commercial purposes. You own all rights to the content you create. The only restriction: make sure your prompt does not include copyrighted third-party material. You are responsible for respecting the rights to any input data.
The model creates images from text descriptions and processes existing pictures: these are TEXT_TO_IMAGE and IMAGE_TO_IMAGE modes. It is suitable for illustrations, covers, concept art, and variations based on a reference. Everything opens in one BotHub window — without VPN or foreign cards, with payment via Russian cards.
The exact resolution and file format are not fixed in our data: you get a finished image as output, not text. The model works both as a generator from a description and as an editor for a submitted picture. It is more convenient to check the parameters of a specific request in the BotHub interface during generation.
There is no specific note about reasoning before answering in our data, so we cannot confirm this capability. Step-by-step reasoning is not required for image generation: the model builds the image immediately from the description. It is easiest to clarify current capabilities in practice.
Function calling and structured JSON output are not noted in our data: these are tools for text models, while here the result is an image. If you need JSON or code integration, BotHub has text models with a unified API, and you can switch without rewriting the integration.
Images — yes: you can input them, and the model will process the picture according to your description. Audio and video are not noted as supported in our data. So the working scenario is a text description plus a reference image, if needed.
We do not have data on prompt and output languages, so we will not promise flawless understanding of Russian. The BotHub interface is in Russian, and it is better to formulate the scene description specifically: object, style, composition. Evaluate the result by the image and refine the request if necessary.