Methods and system for generating an image for a user prompt provided by a user includes receiving the user prompt from the user. The user prompt is analyzed to detect text input and keywords included within. Keyword variations are provided to one or more keywords for user selection. An adjusted user prompt is generated by replacing keywords in the user prompt with keyword variations selected by the user. An image customized for the adjusted user prompt is generated to include image features that are influenced by content of the adjusted user prompt. The customized image providing a visual representation of the adjusted prompt is returned to a client device for rendering.