文字生成图片

Text to image is the most basic task. Describe a scene in words and the model creates a single still image for you.

Steps

1. Open the Image screen from the sidebar.
2. Describe the image you want in the prompt. Naming the subject, setting, and mood together makes the result more reliable.
3. Choose a model, an aspect ratio, and other options.
4. (Optional) Attach a reference image, or use the style chips to tune the mood.
5. Press Generate to run.

Models and options

Changing the model automatically updates the available options — aspect ratios, reference types, and so on — to match that model. The same prompt produces a different style and level of detail per model, so it's worth switching until you find the feel you want.

Adjusting the style

The style and tone chips below the prompt let you tune the mood without rewriting the prompt. Handy when you want to shift the overall tone with a single tap.

Progress and results

Generation takes a moment, and progress is shown. When it's done, the result is added to the gallery where you can view it at full size or download it — and reuse a favorite as a reference for your next generation.

Good to know

• Credits are charged when you generate — the amount depends on the model and options.
• There are limits on how many generations can run at once and per day.
• If a generation goes against the content policy, it's declined and the credits are refunded automatically.