​

Krea Image

Professional AI image generation tool with access to multiple cutting-edge models. Generate images from text prompts with advanced style control and customization options.

Try Krea Image

Access all image generation models from one interface. Choose the perfect model for your creative needs.

Start Creating

​

How it works

Using Krea to generate images couldn’t be simpler. Drop into image mode using the icon in the top navigation, and you’ll see the screen below.

​

1. Choosing a model

In the bottom-left, you can choose which image model you’d like to use. The default option is Krea 1, a model we developed ourselves to specfically create stylish, photographic-style images that avoid a ‘cheesy AI’ look. It’s a great place to start! But you can use the pop-up to access a range of other powerful models, including those you might have heard about elsewhere – like Google’s Nano Banana or OpenAI’s ChatGPT’s Image Model. Each have their own strengths and weaknesses, which you’ll

​

2. Writing your prompt

In the text box, describe the image you’d like to generate. You can be as vague or specific as you like – and the art of prompting is still being refined. We’ll offer our own prompting tips and tricks here in our docs, soon.

​

3. Customize your settings

Most models offer some additional settings – although no model offers all of the settings! These let you tweak your output in some powerful ways. We’ll explain each of those below, in turn.

​

4. Generate

Click generate to create your images! You’ll need to wait anywhere from 10 to 60 seconds, depending on the model you’ve chosen.

​

The generation space

Once you’ve generated your first images, you’ll see your results. It looks like this: Each prompt generates a row of results. (Note, some models will generate four images per request, while others generate two.)

​

Further actions

Click any image to see it in glorious full-sreen. Hover over an image (or set of images) to reveal some extra controls, like so. These let you take further useful actions. Below the images, you’ll see these options:

• Retry - immediately run this hame prompt again, with the same model
• Reuse parameters - brings the prompt, model and settings from that generation into the prompt box, ready for editing and re-running
• Share parameters - generates a link you can copy and send to a colleague, which they can use to generate an image with the exact same prompt and setings
• Download all - downloads all four (or two) images from that generation, in a .zip file.
• Publish - publishes these images to a public page with a link you can share. (A modal pop-up will allow you to add further images, or remove one you don’t included, before the page is pulished.)
• Delete - removes these images from the session view, but NOT from your assets.

Floating above each image on hover, you’ll see:

• Upscale - creates a single, simple upscale of this image in the session chat
• Vary - rapidly generates four similar images
• Video - brings this image into the Video tool as a starting frame for a video generation
• - downloads the image to your device
• - marks the image as a favorite, for easy retreival in the Assets view
• - let us know if the image was especially good or bad

Finally, clicking the ellipsis in the top-right of each image reveals some additional options:

• Edit - brings this image into the Editor tool for AI-powered editing
• Upscale & enhance - brings this image into the Enhancer tool for custom enhancement

​

Working in sessions

The prompt box has now moved to the bottom of the page, with your previous prompt still in place. That means you can, for example, quickly change model or make a small tweak to your prompt, then generate generate another set of images. You can start generating your next set of images while We’ve made it as easy as possible to rapidly experiment, iterate, and explore different directions. On the left-hand side of the page, you’ll see some small thumbnails. Each of these represents a different ‘session’ or conversation. So if you’re working on a new project, starting a new sesssion can help keep things organised. Click the ➕ icon at the top to clear the decks and start a new chat.

​

Image generation settings, explained

When generating images, you’ll find some setting controls beneath the prompt box. These provide you additional over the images you generate. Note, the available settings differ by the model you choose to use. Let’s explore all the available settings.

​

Aspect ratio

Nearly all our models allow you to control the aspect ratio – look for the button that reads “1:1” or similar. Some models limit your selection to a few pre-defined dimensions (typically square, portrait, and landscape) while others allow you to set a custom aspect ratio.

​

Style

The Style function applies a pre-defined visual aesthetic to your image. With styles, you don’t need to describe your desired visual look with your text prompt – the model will automatically try and match the vibe of the provided style. For example, this Urban Crayon style will cause your output to have thick lines, bright simple colouring and cityscape elements: In the Styles pop-up, you can browse through styles that we’ve created, as well as styles shared by members of the Krea community. This is also where you can apply custom styles that you’ve trained yourself. (Read more about training your own model.) Styles are optimised to work best with a particular model. If there’s a style you particularly like, you can click the ‘pin’ icon in the top-right of its thumbnail for ease of retrieval - find them again under the ‘Pinned’ category at the top. Once you’ve chosen your style, you can drag up and down on the corresponding icon, next to your prompt text, to adjust the strength of the effect. The default value is 0.8 out of 1.

​

Style transfer

Similar to the Styles setting, this applies “the look” of a single image to your prompt. This is a great way to generate images that match your existing visual assets. In the pop-up, you’ll be able to choose from a handful of starter images we’ve provided, upload your own image, or point to an image you’ve previously generated on Krea.

​

Image prompt

Add an image to influence your generation. Effects vary based on the model. In some models, aspects of the image prompt – like the characters, style, colour and composition – are combined with the elements of your text prompt to produce the final output. (Imagine copying and pasting a description of the uploaded image, and adding that to your exising text.) The results can be creative, surprising, and a little chaotic – but a great way to come up with something unexpected. In the example below, we use an illustration of a boy skateboarding and a photograph of a white horse in an urban street scene as two image prompts. The resulting image features a boy, skateboard, and aspects of Image A’s illustration style, plus the white horse and background taxis of Image B. In newer ‘reasoning’ models, the AI understands references to the image provided and can execute more specific changes. For example, Google’s Nano Banana can understand a request like “Create an editorial photograph of the woman in the attached image, wearing the red dress in the other attached image.” You can add multiple images – each model has its own limit, from two image prompts (in Krea 1, for example) to 15 (!!) in Nano Banana. Drag up and down on the image thumbnail to adjust the relative influence of the image prompt on the final result. You can add text alongside your image prompt, or simply leave the text field blank. (Prompting with a single image and no text typically creates subtle variations of the original image.)

​

Paint

This setting, available in just two models, allows you to roughly sketch the layout of an image, which will influence the composition of the final output – although not with pixel perfect precision. It can be helpful if you’re trying to drive the AI towards an unconvetional layout (e.g: with the subject unusually small or asymmetrically placed.)

​

Resolution

Some models allow you to generate images at 1.5x, 2x or even 4x size - up to 30 megapixels!

​

Raw

A simple on/off setting that reduces models’ prior aesthetic preferences. All models receive a degree of fine-tuning to generate pleasing images. For example, our own Krea 1 model is optimised to create “cool-looking photos” by default. Selecting ‘raw’ mode removes some of that tendency, making the outputs more ‘boring’ and less artistic by default. That might not sound particularly useful, but it can help if you are trying to generate a particular style of image, and find the ingrained ‘style of the model’ is working against you. In short: if you are trying to achieve a particular look using your text or image prompt, try generating images in both regular and ‘raw’ mode and compare the results.

​

Character reference

Use this tool to define what certain ‘characters’ look like, and then reference them in the prompt. You can upload a single photo for up to three characters, then give each one a name, like @Amy. Then you can write a text prompt like “A photo of @Amy drinking coffee in an Italian piazza”, and the relevant character will appear in your image.

​

Preset

A unique setting for OpenAI’s ChatGPT model, based around some interesting opportunities we’ve found!

​

Choosing the right image model for you

In the bottom-left of each page, you’ll find the model picker. The most recent and popular models are listed first, but you can click View All Models → to see other possibilities.

• Images per generation: Krea 1, Flux, and Flux.1 Krea generate four images per prompt. All the other models generate two.
• **Compute units per generation: **each paid plan comes with a default budget of ‘compute units.’ Each generation uses up a few compute units – but different models consume different amounts. A Krea 1 generation costs just 16 units, while a ChatGPT image generation costs 237 – that’s 14x more! It makes sense to explore and iterate with the ‘cheaper’ models, where possible – but you can always buy add-on compute units if you run out.
• Seconds per generation: some models are faster than others. Krea generates images in about 8 seconds, whereas Nano Banana takes more like 45,
• Resolution: Most model outputs are capped at 1024px square, but Krea 1, Seedream 4, Wan 2.2, Flux.1 Krea and Flux 1.1 Pro Ultra offer higher-resolution outputs – up to 4k in some cases.

Here’s how that looks in a table:

	Images per generation	Compute units per generation	Seconds per generation	Maximum resolution
Krea 1	4	16	8	4096
Flux	4	14	5	1024
Nano Banana	2	114	45	1024
ChatGPT Image	2	237	60	1024
Seedream 4	2	84	20	4096
Wan 2.2	2	30	20	1536
Qwen	2	8	15	1024
Imagen 4	2	114	60	1024
Runway Gen-4	2	140	60	1024
Ideogram 3.0	2	114	30	1024
Flux.1 Krea	4	16	8	4096
Flux Kontext	2	14	5	1024
Flux Kontext Pro	2	114	10	1024
Flux 1.1 Pro	2	114	10	1024
Flux 1.1 Pro Ultra	2	171	30	2048
Imagen 4 Ultra	2	171	60	1024
Imagen 4 Fast	2	57	20	1024
Imagen 3	2	114	60	1024
Seedream 3	2	84	5	1024

As we discussed in the Image generation settings section, models also vary by what other settings they offer. For example, only some models accept Styles. Here’s a breakdown:

	Aspect ratio	Image style	Style	Image prompt	Preset	Paint	Raw	Character reference
Krea 1	✅	✅	✅	✅			✅
Flux	✅		✅	✅
Nano Banana		✔️*		✅		✅		✔️*
ChatGPT Image	✅	✔️*	✅	✅	✅	✅		✔️*
Seedream 4	✅	✔️*		✅				✔️*
Wan 2.2	✅		✅				✅
Qwen	✅		✅				✅
Imagen 4	✅
Runway Gen-4								✅
Ideogram 3.0	✅			✅				✅
Flux.1 Krea	✅	✅	✅	✅			✅
Flux Kontext	✅			✅
Flux Kontext Pro	✅			✅
Flux 1.1 Pro	✅
Flux 1.1 Pro Ultra	✅
Imagen 4 Ultra	✅
Imagen 4 Fast	✅
Imagen 3	✅
Seedream 3	✅

*Note: Seedream 4, Nano Banana and ChatGPT Image are reasoning models with intelligent understanding of prompts. Although they do not offer ‘character reference’ or ‘style reference’ as explicit settings, you can upload images of characters or artwork styles, and simply write a prompt that says "Create an illustration of a skyscraper in the style of the attached image."

​

Krea Models (Recommended)

Krea 1

Our frontier model that solves the “AI look” problem with realistic textures and diverse styles.

Flux Family

Fast, versatile generation with extensive style support and custom ratios.

​

Third-Party Models

We also provide access to leading models from top AI providers:

Imagen 4

Google’s latest model with superior typography and photorealistic detail.

ChatGPT Image

OpenAI’s model with advanced reasoning and complex scene understanding.

Ideogram

Aesthetic models specialized in graphic design and flat art styles.

Runway Gen-4

Cinematic image model optimized for character compositions.

​

Model Comparison

Model	Speed	Best For	Key Strength
Krea 1	Few seconds	Photorealism	Solves “AI look” problem
Flux	Few seconds	General use	Style references
Nano Banana	~60 seconds	Prompt adherence	Reasoning, world knowledge, complex prompts
ChatGPT Image	~60 seconds	Complex scenes	Advanced reasoning
Ideogram	~30 seconds	Graphics	Flat, aesthetic styles, typography
Runway Gen-4	~60 seconds	Characters	Cinematic quality

​

Getting Started

​

For Beginners

1. Start with Krea 1 for the most realistic results
2. Try simple prompts like “a photo of a sunset over mountains”
3. Experiment with style references using your own images
4. Use preset styles to explore different artistic directions

​

For Advanced Users

1. Compare Flux variants for different speed/quality needs
2. Use Imagen 4 when you need text in your images
3. Try ChatGPT Image for complex, multi-element scenes
4. Leverage prompt engineering techniques for precise control

​

Use Cases

Content Creation

• Social media visuals and marketing assets
• Blog illustrations and hero images
• Product photography alternatives

Creative Projects

• Concept art and storyboarding
• Character design and world building
• Digital art and illustrations

Professional Work

• Brand assets and marketing materials
• Website graphics and UI elements
• Print design and editorial imagery

​

Tips for Best Results

​

Prompt Writing

• Be specific about style, lighting, and composition
• Include technical terms for photorealistic results
• Use artist names or art movements for style direction
• Combine abstract concepts with concrete details

​

Model Selection

• Photorealism: Krea 1 or Imagen 4
• Speed: Krea 1 or Flux variants
• Typography: Imagen 4
• Artistic styles: Flux or Ideogram
• Complex scenes: ChatGPT Image

​

Style References

• Use high-quality reference images
• Match reference style to your desired output
• Adjust influence strength with sliders
• Combine multiple references for unique looks

Krea Image brings together the best AI models in one professional interface, giving you the power to create exactly what you envision.