Describe the subject, composition and mood, then generate an image and review it before saving. Follow the real example below, then adapt it to your own source. Open AI Image Generator.
Write the visual brief
Open AI Image Generator and enter the greenhouse prompt shown in the comparison. It specifies a subject, setting, light and composition, with no lettering requested. The exact wording is available below.
Prompt limit: 4,000 characters and 16,000 UTF-8 bytes. One image per request at one of the three listed canvas sizes.
Use the exact example input
A small glass greenhouse in a sunlit meadow, soft morning light, clean composition, no lettering
Prompting creates an image. To put exact wording on a photo you already have, use Add Text to Image.
Choose the canvas and settings
The recorded example uses a square canvas and the settings below. Optional Style, Artist, Quality, Color and Mood choices can add descriptive guidance; the example uses the written brief without additional style choices.
Seed accepts 1 to 100,000,000 and has a randomize action. Detail steps accepts 1 to 60, while Prompt guidance accepts 1 to 15. The initial settings are seed 1, 30 steps and guidance 7. Changing them can affect variation, processing time and how closely the prompt is followed; a higher value is not always better.
The example in this guide
Verified in the i2TEXT browser workflow on November 11, 2026.
- Input
- prompt.txt
- Output
- generated-1.png
- Saved result
- 1449606 bytes
Generate once, then inspect
Select Generate to submit the request. Inspect the returned picture for composition, unwanted details and its suitability for the intended use. The after image is the PNG actually downloaded from this run.
Another request can produce a different result. The example records one output rather than promising a pixel-identical image from the same prompt.
Save the picture
Download generated-1.png before refreshing or leaving the result. The recorded PNG is 1024 by 1024 pixels.
Generate sends the prompt and selected options for AI processing. Inspect the result for unwanted details, then download the PNG. Image options can prepare supported alternative image formats. Continue to can hand the image to Image to Text or Add Text to Image without starting their processing automatically.
Written promptgenerated-1.png
Common questions
How large is the generated image?
Square produces 1024 by 1024 pixels. Portrait produces 768 by 1024, and landscape 1024 by 768. Choose the shape that suits the intended use before generating; each request returns one image.
What do the advanced settings change?
Seed accepts 1 to 100,000,000 and has a randomize action. Detail steps accepts 1 to 60, while Prompt guidance accepts 1 to 15. The initial settings are seed 1, 30 steps and guidance 7. Changing them can affect variation, processing time and how closely the prompt is followed; a higher value is not always better.
Can I rely on exact lettering in the generated picture?
No. Review generated lettering carefully. For an exact caption, open the picture in Add Text to Image and place an editable text layer before exporting.
Try AI Image Generator with your own source.
Use your own source and review the result before sharing.
