How to Generate an AI Image on OpenArt: Step-by-Step Guide

Generating an AI image on OpenArt comes down to four decisions: which model to use, how you write the prompt, which settings to dial in, and what to do with the result once it's on screen. This guide walks through each step in order, starting from the homepage and ending with a finished image ready to download, edit, or turn into a video. Watch the full walkthrough below, or skip straight to the written steps.

Opening Create Image from the homepage

On the OpenArt homepage, select Create, then choose Create Image to open the AI Image Generator. That's the page where every decision below happens, starting with the model.

Choosing a model: prompt-only vs. reference

Once the Create Image page is open, the first decision is which model to use, and that comes down to whether text alone can describe the image or an uploaded image needs to guide it. OpenArt selects a model automatically, but the model picker gives you control over that choice.

Prompt-only models generate images solely from written instructions. Recraft V4 is one example. Choose this type when you do not have a reference image or when your prompt can describe the subject, setting, and style clearly enough.

Reference models accept uploaded images as visual guidance. Nano Banana 2 can use a reference to preserve details around a face or product. It can also follow a pose or visual style more closely across generations. Choose a reference model when consistency matters, especially when creating several images of the same character or item.

Model picker filters can narrow the options by capabilities such as reference support or 4K output. Apply the reference filter when you plan to upload an image. Apply the 4K filter when you need a larger output, then confirm the available resolution in the selected model's settings. Model capabilities can change, so check the model dropdown rather than assuming every model offers the same controls.

Writing the prompt: OpenArt's structure for subject, style, and everything between

With a model selected, the next step is the prompt itself. If you are learning how to write AI prompts, OpenArt's structure gives you a repeatable way to solve the blinking cursor problem. You build the scene one layer at a time instead of trying to describe the finished image immediately. OpenArt recommends starting with subject and action, then adding subject details, environment and lighting, shot type, and style.

Your first sentence should name the subject and describe an action. A useful first draft can be as simple as "A man walking in the city." Avoid openings such as "I want a beautiful picture of" or "I would like to see." Those phrases do not tell the model what should appear in the image.

Specific adjectives help the model define the subject. You might expand the prompt to "A middle-aged man in a charcoal overcoat walking in the city." Details such as age, clothing, expression, and physical appearance give the model visible information to interpret. Add only details that affect the image you want.

Environment and lighting establish the setting and mood. The next version might read "A middle-aged man in a charcoal overcoat walking through a busy city street at dusk, with warm storefront light reflecting on wet pavement." The environment identifies where the action happens, while the lighting describes how the scene should look.

Shot type and style shape the final presentation. Put the shot type at the beginning so the model knows how to frame the subject, and place the visual style at the end. The completed prompt could read "A wide shot of a middle-aged man in a charcoal overcoat walking through a busy city street at dusk, with warm storefront light reflecting on wet pavement, cinematic editorial photography."

The finished order gives you a practical template.

Shot type, subject with adjectives, environment and lighting, style.

Useful AI image generator prompts describe choices the model can represent visually. "Wide shot" affects composition, "charcoal overcoat" affects appearance, and "warm storefront light" affects color and illumination. Each phrase has a specific job.

OpenArt treats this structure as a scaffold rather than a fixed formula. Once you feel comfortable, you can reorder elements, remove details, or test different styles to see how your chosen model responds. The structure provides a reliable starting point when the prompt box feels empty, but experimentation helps you learn which wording produces the images you want.

Dialing in generation settings

Below the prompt, the settings control everything about the output. Resolution determines how many pixels OpenArt generates. Choose a higher resolution when you need more detail for print, cropping, or a large display. GPT Image 2 supports 4K output, as do Nano Banana 2 and Nano Banana Pro. OpenArt updates model capabilities frequently, so confirm 4K support in the selected model's dropdown. If you're deciding between models, this GPT Image 2 vs. Nano Banana 2 comparison breaks down how their outputs differ.

Quality controls appear only when the chosen model exposes them. For example, GPT Image 2 lets you adjust quality, while other models use a fixed setting. Higher quality may suit a final asset, while a lower setting can help with early prompt testing.

Aspect ratio controls the canvas shape. Match it to the image's destination before generating. A square ratio fits many feed posts, while a wide ratio works better for banners or presentation slides. Choosing the intended shape now reduces later cropping.

Quantity controls how many takes OpenArt creates from one prompt. Generate one image when you want to test a small prompt change. Choose several when you want alternative compositions and details without submitting the same prompt repeatedly.

Using Prompt from Image, Enhance Prompt, and Clear Prompt

Don't want to write a prompt from scratch? Image to Prompt converts an uploaded image into an editable text prompt. The uploaded image helps describe elements such as the subject, setting, lighting, and style in words. You can revise the generated prompt before creating an image.

Prompt from Image differs from the reference-model workflow covered earlier. A reference model uses the uploaded image as visual guidance during generation. Prompt from Image only writes prompt text, so the source image does not directly guide the final output.

Enhance Prompt refines the text already in the prompt field by adding detail or making the instructions more specific. Review the revised version before generating to make sure it preserves your intended subject and style. Clear Prompt removes all text from the field when you want to start again.

What happens after you hit Create

With the model, prompt, and settings in place, selecting Create is the last step. OpenArt places the generated images in a results grid. Review each result, then download the image you want to keep or choose Send to Edit Image to open it in the AI Photo Editor for further changes.

Choose Send to Video when you want to use a generated image as the starting point for a video with the Image to Video feature. OpenArt treats Image Variations as a separate tool for creating multiple versions of an existing image, so you will not find it within the initial generation flow.

FAQs

A few questions come up often once people start working through this flow.

What if I do not have a reference image?

A prompt-only model creates images using text instructions alone. OpenArt offers prompt-only models such as Recraft V4. You can generate an image without uploading any visual material.

Do I need to follow the recommended prompt structure exactly?

The prompt structure provides a flexible order for describing the shot, subject, environment, lighting, and style. OpenArt recommends it as a starting point rather than a fixed rule. You can change the order or omit details as you learn what produces the results you want, and best AI image generator prompts has more examples to borrow from.

Which OpenArt models support 4K output?

4K output generates images at a higher pixel resolution than standard options. OpenArt lists GPT Image 2, Nano Banana 2, and Nano Banana Pro among the models that support 4K. Check the model picker before generating because available models and settings can change.

Can I generate more than one image at once?

The quantity setting controls how many images OpenArt creates from one prompt. You can increase the quantity in the generation settings panel before selecting Create. Multiple results let you compare different takes without submitting the prompt repeatedly.

Closing takeaway

That's the full path from opening Create Image to a finished result. Prompting improves when you compare results and revise specific details. Use OpenArt's structure as a starting point, then change one element at a time, such as lighting or shot type, to see how each phrase affects the image. If you want inspiration for scroll-stopping portraits, browse best Nano Banana prompts for viral photos.

Try the workflow in the AI Image Generator and refine your prompt based on the results.