Generating images

You can use Theta Assist to generate images. Just tell your assistant what you would like to see in your image, any style requirements, and see what the AI comes up with.

This can be handy for visualising your ideas, generating images for presentations, generating marketing content, etc.

Plan limitation: Image generation not available when using your own Azure OpenAI service.

Quick links

  • Image settings
  • How to download generated images
  • Tips and tricks
    • Start from scratch
    • Use thread files as inputs
    • Human spell-check
    • Prompting tips
  • Supported models
  • Plan limitations

Image settings

Choose the model, size, quality and complexity for generated images, and whether to show progress previews.

  1. Open image settings. Select the wrench icon below the message box, then open the image generation tab.
  2. Choose a model. Use the default unless you need a different image generation model.
  3. Choose a size. Select the output dimensions, or choose Auto to let the model select the best fit for your prompt.
  4. Choose a quality level. Higher quality improves image detail but uses more tokens.
  5. Choose a complexity level. This controls how much planning goes into the image before it is generated. Higher complexity can improve how elements are arranged, but takes longer and uses more tokens.
  6. Set progress previews. Turn Progress previews on to see preview images during generation. Turn it off to use fewer tokens.
  7. Select Close. Your settings apply to the current thread.

Image settings reset to the default when you start a new thread.

Wrench icon used to open thread settings.

Image generation defaults

How to download generated images

To download an image, click the image download button below the image:

 

image download

 

Tips and tricks

Start from scratch

Any images already generated in a thread are remembered by the assistant. If you want to start from scratch, start a new thread. Otherwise you may see elements from previous images appearing in new ones.

Use thread files as inputs

Want to incorporate a logo or other visual element into your generated image? Attach the logo as a thread file.

Human spell-check

Carefully check any text in a generated image. AI is more likely to make mistakes with text in images than in text responses.

Prompting tips

Image generation works best when you use terms like "draw" or "edit" in your prompt. For example, if you want to combine images, instead of saying "combine" or "merge", you can say something like "edit the first image by adding this element from the second image".

 

Supported models

The image generation tool is supported by most recent models. To check, look for the image generation badge when switching models.

 

Plan limitations

Image input is now supported when using your own Azure OpenAI service (maximum 10 images per thread). 

Image generation is not supported when using your own Azure OpenAI service.