An AI image generator interprets a written description to create a new picture. On this page, you start with text rather than an uploaded image. For edits based on an existing picture, or work guided by reference images, use Image to Image. Those are separate inputs and workflows.

Match the model's available aspect ratios, resolution and formats to where you plan to use the image. To compare visual results, try the same clear prompt across models and keep output settings comparable where possible. The linked model pages explain each model's specific capabilities.

Use this structure: subject, setting, composition, style or lighting, then any required text or elements to avoid. For example: A white ceramic mug on a light wooden table, placed on the left with empty space on the right. Soft window light, realistic product photography style, no text or logos. This teaching example is separate from the poster prompt above and has no paired sample image. Aim for clear instructions, not maximum length; the input limit varies by model.

You can write prompts in English, Chinese, Japanese, Spanish or Portuguese; the poster example uses Japanese. Describe the subject and style in your preferred language. For words that should appear inside the image, specify their language and exact wording.

This page does not currently provide a separate negative prompt field. Include exclusions in the main prompt, such as no text or logos, and describe what you want to see, for example, a plain, unbranded mug. Unwanted details may still appear.

Start with the intended use: a vertical poster, a wide cover or a square post. Choose a matching aspect ratio from the available options, then set the resolution and quality offered by the model. Writing 4K or 8K in the prompt does not override those output settings.

Exact text, spelling, logos and layout are not guaranteed. Put the desired wording in quotation marks and describe its placement, then inspect every letter and symbol in the result. For a design that needs exact typography or branding, finish those details in a suitable editing or design tool. This page does not provide editable text layers.

Repeating a prompt can produce a different image or change a character's appearance. This page does not offer a public seed control for reproducing results. If you want to work from an existing output, take it to Image to Image as a reference and describe what should stay or change.

Output choices depend on the selected model and may include PNG, JPEG or WebP. Some models offer a transparent background setting; choosing PNG alone does not make a background transparent. Check the model's controls, or use Background Remover afterward if needed. The page does not offer SVG, PSD or layered design-file export.

Generation requires signing in and uses credits. Available bonus credits can be used when your balance covers the task. The displayed credit estimate reflects the selected model and settings. Each new generation, including another attempt with a revised prompt, uses credits. You can review charges and adjustments in your credit history.

Generation time varies by task; follow its status for progress. If a task fails, use the error message to decide what to change before retrying. If it finishes but the image does not match your idea, make the relevant subject, composition or style instructions clearer and generate a new version.

Use Image to Image to edit an existing picture, Image Upscaler to enlarge it, or Image to Video to create a clip from it. For background or watermark removal, open the corresponding tool. These tools run separately, with their own input requirements and credit rules.