Help·Generate an image from text
Generate an image from text
Text-to-image
Describe it, generate it
Text-to-image generation lets you create images by writing a description — called a prompt. You describe what you want to see, choose a model, and the AI generates an image based on your words.
AI generation is available to subscribers on the Dali tier ($10/mo) and above. Each tier comes with a monthly credit allowance.
How to generate an image
Step by step
1. Go to the Generate page
Navigate to Generate from the navigation menu. Make sure the Image tab is selected.
2. Write your prompt
Describe the image you want to create. Be specific about subject, style, lighting, and mood. For example: "a neon-lit Tokyo alleyway at night, rain-soaked streets, cinematic photography" will produce very different results from "Tokyo street".
3. Choose your model
Select a model from the dropdown. Flux Schnell is the fastest for quick experiments. WAN 2.1 14B and Fooocus are great for detailed, artistic results. See What are AI models? for a full comparison.
4. (Optional) Add LoRAs
If the model supports LoRAs, you can add up to 3 to steer the style. LoRAs are small style modifiers — like "anime style" or "film grain". See What are LoRAs? to learn more.
5. (Optional) Adjust settings
Open advanced settings to control image size (1
, 16, 9, etc.), inference steps (more steps = more detail but slower), and guidance scale (how closely the model follows your prompt).6. Generate
Tap Generate. Your image will appear in the queue and be delivered to your Inbox once complete. Fast models take 2–4 seconds; detailed models can take up to 20 seconds.
Writing better prompts
Get the results you're imagining
The quality of your prompt directly affects the quality of your result. Here are some tips:
Be specific
Instead of "a cat", try "a ginger tabby cat sitting on a windowsill, golden hour light, shallow depth of field, 35mm photography".
Include style direction
Mention the medium or style you want: "oil painting", "cinematic photograph", "watercolour illustration", "3D render".
Use negative prompts
If the model supports it, add a negative prompt to exclude things you don't want — like "blurry, low quality, text, watermark".
Generate an image from another image
Use an existing image as a starting point for your generation.
Generate a video from an image
Turn your generated images into short animated videos.