Vove AICreate
Back to blog

Create an AI image from a prompt or reference

A practical guide to prompts, image references, output settings and reviewing your generated image.

Oct 30, 2026

Start in the image workspace with a clear description of the image you need. If the subject already exists, add a reference image to guide its appearance. A reference helps communicate your intent, but the generated result can still change labels, faces and shapes.

Write the scene before choosing a style

Name the subject, its placement and the intended use. “A ceramic mug on a light wooden kitchen table, morning window light, handle visible, empty space on the right for a headline” gives the model more direction than “beautiful product photo.”

Add the details that must stay consistent. For a product, that might be its silhouette, material and label. Avoid packing several unrelated scenes into one request.

Add a useful reference

Choose a clear PNG, JPEG or WebP within the upload limit shown in the interface. Keep the important subject visible. You can also apply a reference example or a skill brief, then replace the reference with your own image.

Reference examples are starting points for composition and style. They are not a guarantee of the result you will receive, or a claim that a named customer made them.

Review settings and credits

Choose a model, aspect ratio and supported resolution. The available options follow the selected model. Check the credit estimate before pressing Generate. If you are asked to sign in, the pending creative inputs are kept so the requested operation can continue afterward.

Generation uses credits. Do not assume that a higher resolution will recover details missing from a source image; a model may invent them.

Inspect the output

Find your own tasks in Assets and Creation history. Compare the whole result with your source, especially text, logos, edges and product details. Change one instruction at a time if you want a different result.

For manual image-and-text composition, add your image to Canvas and export a PNG. Canvas is a separate composition tool; it does not turn generated images into editable layers.

Common questions

Can I use an image without uploading a reference?

Yes. A prompt can start a text-to-image task. Some specialist editing pages require an image because their purpose is to transform an existing source.

Will the model preserve exact text?

There is no guarantee. Keep text short, proofread every result and add exact copy manually when accuracy matters.

Where should I ask about an account problem?

Open Contact for the available support channels. Do not share passwords or verification codes.

Before your next image

Keep the source, write one concrete change and check the estimate. Review the downloaded image at the size you intend to use it.