Skip to main content
Updated over a week ago

generate image with prompts


  • Keywords that describe the image you want to make. It could be phrases or words. Separate with comma.

  • Describe with Tags: PixAI's models work best when you word your prompts in the form of descriptive 'tags' or short phrases. You'd want to separate your tags with commas. Here's a basic example of how tags are worded:

  • Emphasis: Tell the AI what tags you wish it to focus on by surround tags with '()' --or in multiple: '(())'-- to strengthen it. Or you could do so by weakening a tag with '[]'.

  • The order of your prompt matters: Try to describe the main object/character at the beginning of your prompt. For example, for character portraits, our users like to start with '1girl' or '1boy' as their first tag.

  • Ratios: The usage of ratios or weights with our models does not always produce the most stable or obvious results. Please keep this in mind while experimenting with your prompts.

  • Notes

    • In PixAI, you can use the Note function to save the setting template which can memorise prompts, negative prompts, models, size, sampling steps and methods, base image. Once the note is saved, you can use it as a template any time by importing it.

generate image with base image

Image to Image

  • Upload Image

    • Upload an image as a reference

  • Paint

    • Edit uploaded image and save its edited result as image reference

  • Describe an Image

    • Describe the image uploaded as reference by AI to get prompt description.

Image Edit


  • You can use scroll up and down to control the whole canvas size.

  • Switch mode between Inpaint and Outpaint will discard all changes.

  • InPaint

  • OutPaint

  • Reset

    • Discard all changes in each mode.

generation result

  • Publish

    • You can publish your task to PixAI community once you fill in the post details.

  • Describe

    • You can get generated prompts from selected task.

  • Upscale/Enhance

    • Upsacle allows you to increase the size of your image without altering it.

    • Enhace will generate a new, larger, detailed image based on the original.

  • Variation

    • Variation will take the current image as base image generate 3 new different images using the same prompts and settings with the base image task.

  • Edit

    • Edit will bring the task to "Edit Image" page

  • Use as Base Image

    • Take the selected image as reference image

  • Download

    • Download this image as a PNG file


  • Model

    • Your artwork model. Different models have different styles, such as photo-like, realistic, anime and so on.

    • You can find more models from our Model/LoRA market by clicking "Browse Model Market"

  • LoRA

    • Expanded learning packet for model. It can make generation more targeted or detailed on specific character/style. It can let AI be able to recognize specific prompts.

    • A LoRA can make AI recognize specific prompts for a better result.

    • You can find LoRAs in PixAI Model/LoRA Market

  • HiRes

    • HiRes means high resolution. If you check this box, you can set a larger aspect ratio. Choosing HiRes will cost more credits and generate more time.

  • ControlNet

    • ControlNet can analyse the image you upload to shape the frame of your generation.

  • Composition

    • Composition allows you to create images that specified details like in which part of the canvas will have what content.

  • Negative prompts

    • Negative prompts tell the AI what to avoid. On our generation page, we provide a list of default negative tags to filter out things that might lower the quality of your image. You are welcome to change or add other undesired content to the list.

  • Sampling Steps

    • Steps define the number of iterations the AI should refine from the initial creation of your prompt. Higher steps will cost more credits.

  • CFG Scale

    • The scale value decides how closely the AI follows your prompts. Also, when the scale is low, AI tends to produce softer, painterly pictures. We recommend keeping the scale below 7.

Did this answer your question?