AITiny can edit existing images using a text prompt — a technique known as image-to-image (or img2img) editing. You describe the changes you want to make, and the AI produces a new version of the image based on your instructions while preserving the rest.
Image editing uses the same image service configuration as image generation. Not all image services and models support image editing. When the configured service does not support it, the context menu option and the button in media fields will not appear.
Services that support image editing: OpenAI (GPT Image models), Google Gemini, and Black Forest Labs Flux. Services that do not support image editing: Google Imagen and Automatic1111.
The image editing feature is available in two places:
Right-click context menu in TinyMCE and JCE: right-click on any image inside the editor. If the configured image service supports image editing, an option will appear in the context menu. Selecting it opens the image editing modal with the clicked image loaded as the reference.
AI Edit button in Joomla media fields: when a media field (such as Intro Image or Full Article Image) already has an image selected and the configured image service supports image editing, an button appears next to the field. Clicking it opens the image editing modal using that field's image as the reference.
The image editing modal works like the image generation modal, with the current image displayed as a reference. The controls are:
Describe the changes you want to make to the image. Be specific about what to change and what to keep. For example: "Change the background to a sunset sky, keep the subject unchanged" or "Convert to a watercolour painting style".
Uses the text AI service to suggest an edit prompt based on the article content. Useful when you want a contextually relevant edit but are unsure what to ask for.
An optional image style preset. When selected, the style prompt is combined with your edit prompt. Not all services apply style presets during image-to-image editing operations.
These options are not used for image editing. The output dimensions are determined by the source image and the capabilities of the selected image service.
Click to start the editing process. When the edited image is ready, you can:
Click to save the edited image. It will be saved to the configured media adapter and subdirectory, and the image in the editor or media field will be updated to show the new version.
Click to generate a new variation with the same prompt.
Adjust your prompt and click again to try a different edit.
Click (or press Escape) to cancel without making any changes.
Be explicit about what to change and what to keep. Vague prompts like "make it better" rarely produce useful results. Something like "Replace the white background with an outdoor cafe scene, keep the person and pose unchanged" gives the AI clear guidance.
Image editing works best with clear, high-resolution source images. Heavily compressed or very small images may produce lower-quality edits.
Different services handle image editing differently. OpenAI GPT Image models excel at targeted edits while faithfully preserving the rest of the image. Google Gemini understands context and scene composition well. Flux Kontext models are strong at style transfer and precise localised changes.
If the AI changes too much or too little, refine your prompt. Phrases like "subtle adjustment" or "only change the background" can help control the extent of the edit.
Like image generation, editing is non-deterministic: the same prompt on the same image can yield different results on successive runs. Use to explore variations.
Image editing shares the same service configuration as image generation. The only additional parameter specific to image editing is the AI Edit Icon, which controls the icon displayed on the button in media fields. See the Image generation parameters section for the complete list of configuration options.