# MiniMax H3 Image MiniMax H3 Image is MiniMax's open-weight image model. In Sloosh it makes 1K, 2K or 4K images (2K by default) from a prompt, or edits and combines up to 16 reference images. Web page: https://sloosh.ai/models/minimax-h3-image [Use MiniMax H3 Image](https://sloosh.ai/studio/image?image-model=minimax-h3-image) ## Key facts in Sloosh | | | | --- | --- | | Made by | MiniMax | | Makes | image | | Inputs | text, up to 16 reference images | | Resolution | 1K, 2K or 4K (default 2K) | | Aspect ratios | 1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 4:5, 5:4, 21:9, 9:21 (default 1:1) | | Prompt limit | 32,000 characters | | Open weights | yes, on Hugging Face under the MiniMax-H3-Image Community License | | Transparent background | no | ## Best for - Images that need to hold up big, rendered at up to 4K. - Edits from several inputs at once: it takes up to 16 reference images, referred to by position (image 1, image 2, ...) in the order you add them ([MiniMax](https://platform.minimax.io/docs/guides/image-generation)). - Wide and tall layouts, from 21:9 banners to 9:21 phone screens. ## How to use MiniMax H3 Image in Sloosh 1. Open [Image Studio with MiniMax H3 Image selected](https://sloosh.ai/studio/image?image-model=minimax-h3-image), or pick MiniMax H3 Image in the model chip. 2. Add references if you have them (up to 16): from the drawer, by pasting, or with Use as reference on an earlier result. 3. Write the prompt. Name each reference by number and job, for example "Image 1: product photo". 4. Set aspect ratio, resolution (2K by default, up to 4K) and how many images. 5. Check the cost on the Generate button, then generate. ## MiniMax H3 Image prompting tips - Write a detailed description in English. The model card recommends detailed English prompts ([Hugging Face](https://huggingface.co/MiniMaxAI/MiniMax-H3-Image)). MiniMax's API can also enhance a prompt before it renders. - Refer to references by their order, Image 1, Image 2 and so on, as they sit in the tray ([MiniMax](https://platform.minimax.io/docs/guides/image-generation)). - Describe what you want to see rather than what to leave out. The model takes no negative prompt ([Hugging Face](https://huggingface.co/MiniMaxAI/MiniMax-H3-Image)). - Work at 2K while you get the look right, then render the keeper at 4K. ## MiniMax H3 Image example prompts **Prompt 1** ```text A close-up product photo of a brushed steel wristwatch with a deep green dial, lying on a slab of dark grey slate. Soft window light from the left, a thin reflection along the bezel, fine scratches visible on the slate. Shallow depth of field, the strap fading out of focus at the back. Horizontal 16:9 frame for a website banner, empty slate on the right for copy. ``` **Prompt 2** ```text A wide 21:9 photo of a quiet harbour town at dawn. Pastel houses along the water, fishing boats tied up in a row, mist sitting low on the bay, gulls on the harbour wall. Natural colour, fine detail in the windows and rigging, no people. ``` **Prompt 3** ```text Image 1 is a linen shirt. Image 2 is the model. Image 3 is a sunlit terrace. Put the shirt from image 1 on the model in image 2, sleeves rolled once, standing on the terrace from image 3. Keep her face, hair and pose exactly as in image 2, and the shirt's colour and buttons exactly as in image 1. ``` ## How it compares in Sloosh | Model | Resolution | References | Prompt | | --- | --- | --- | --- | | MiniMax H3 Image | 1K, 2K, 4K | 16 | 32,000 | | [MiniMax H3 Image Highspeed](https://sloosh.ai/models/minimax-h3-image-highspeed) | 1K, 2K, 4K | 16 | 32,000 | | [GPT Image 2.6 Sunburst](https://sloosh.ai/models/gpt-image-2.6-sunburst) | 1K, 2K, 4K | 16 | 32,000 | | [Nano Banana Pro](https://sloosh.ai/models/gemini-3-pro-image-preview) | 1K, 2K, 4K | 14 | 50,000 | | [Seedream 5 Pro](https://sloosh.ai/models/seedream-5-0-pro) | 1K, 2K | 10 | 5,000 | MiniMax H3 Image Highspeed is MiniMax's Highspeed tier, with the same inputs and limits at a higher price; the Generate button shows it. GPT Image 2.6 Sunburst also makes GIFs, SVGs and transparent PNGs. ## Where else it runs - MiniMax API, models `MiniMax-H3-Image` and `MiniMax-H3-Image-Highspeed`, with text-to-image and image edit endpoints (MiniMax image generation guide, accessed 2026-11-10). - Open weights on Hugging Face, `MiniMaxAI/MiniMax-H3-Image` (accessed 2026-11-10). ## Frequently asked questions ### What is MiniMax H3 Image? MiniMax's open-weight image model. In Sloosh it makes 1K, 2K or 4K images from a prompt and up to 16 references. ### What resolution is MiniMax H3 Image? 1K, 2K or 4K in Sloosh, and 2K is the default. ### How many reference images does MiniMax H3 Image take? Up to 16 per generation in Sloosh, numbered in the order you add them. ### What is the difference between MiniMax H3 Image and Highspeed? [Highspeed](https://sloosh.ai/models/minimax-h3-image-highspeed) is MiniMax's Highspeed tier, with the same inputs and limits at a higher price; the Generate button shows it. ### Is MiniMax H3 Image open source? Its weights are public on Hugging Face under the MiniMax-H3-Image Community License, which has its own terms. In Sloosh you use it with no setup or API key. ### Can MiniMax H3 Image make a transparent background? No. For a transparent PNG, use [GPT Image 2.6 Sunburst](https://sloosh.ai/models/gpt-image-2.6-sunburst). ### Does MiniMax H3 Image have a seed? No. Run it again, or ask for more images in one run, to get variations. ## Related - Models: [MiniMax H3 Image Highspeed](https://sloosh.ai/models/minimax-h3-image-highspeed), [GPT Image 2.6 Sunburst](https://sloosh.ai/models/gpt-image-2.6-sunburst), [Nano Banana Pro](https://sloosh.ai/models/gemini-3-pro-image-preview), [Seedream 5 Pro](https://sloosh.ai/models/seedream-5-0-pro) - Docs: [MiniMax H3 Image](https://sloosh.ai/docs/minimax-h3-image.md) ## Sources - [MiniMax: Image generation guide](https://platform.minimax.io/docs/guides/image-generation) - [MiniMax: Pricing](https://platform.minimax.io/docs/guides/pricing) - [Hugging Face: MiniMax-H3-Image model card](https://huggingface.co/MiniMaxAI/MiniMax-H3-Image)