Image generation

Use the EvoX image-generation plugin to create an image from text or edit an existing image with reference inputs.

Enable the plugin

The built-in plugin starts disabled. Choose a provider in its dedicated configuration and use credentials managed under Settings > AI services. Cancellation, missing credentials, or a failed test keeps the plugin disabled.

The plugin does not silently move to another provider after failure or hide possible cost.

Write an effective prompt

One to three clear sentences are often enough. Include the purpose and audience, subject and action, scene and composition, size, lighting, color, material, required stable elements, and forbidden content.

For edits, say exactly what changes and what remains unchanged. Label multiple references and explain whether each controls content, style, or layout.

Review and refine

Outputs are saved to the task artifact directory with provider, model, dimensions, MIME type, size, and local path. Inspect subjects, hands, text, brand elements, edges, and transparency before requesting one focused edit at a time.

Keep in-image text short and review it character by character. Image generation can draft diagrams and posters, but typography-heavy final work should be finished in a design or web tool.

References, rights, and cost

Input images and masks must be ordinary allowed local files that pass type and size checks. Do not present generated images as evidence of real events, people, or products. Confirm portrait rights, avoid unauthorized brand material, and review provider, inputs, cost, and output location before remote calls.

EvoX Docs · Features · Capabilities