How to generate an image from a text prompt

Text-to-image is the starting point for most of what else on this site does — an image you generate here can be upscaled, edited, animated into video, or used as the reference for a talking avatar. Getting the prompt right the first time is less about clever wording than about being specific.

Step by step

1. 1

   Describe the subject, setting and style together

   A single sentence with all three does better than a vague noun. 'A red apple on a wooden table, studio lighting' outperforms 'an apple'.

2. 2

   Pick a model

   The default suits most prompts. Try a different one if a particular style — photoreal, illustrative — matters more than usual.

3. 3

   Generate a few at once

   Up to four variations in one request, each billed as one image. Comparing variations side by side is faster than iterating on one.

What makes the difference

• State the aspect ratio you actually need up front — cropping a square result down to 16:9 loses more than generating it right the first time.
• Iterate on one change at a time. A prompt that tries to fix three things at once usually fixes none of them cleanly.
• A generated image saves straight to your asset library, ready to feed into image editing, upscaling or video tools without re-uploading.

Questions