Open standard docs
Integration Guide for AI Image Platforms
How image generation platforms can support Open Image Templates while keeping prompts portable and generator-neutral.
Integration goal
An image platform should be able to import an Open Image Template URL, render the editable fields, compile a visible prompt, and generate an image without requiring hidden prompt text or a NanoGPT-specific account.
Template discovery
Support both human pages and JSON endpoints. If a user pastes /templates/{id}, look for the linked /templates/{id}.json endpoint. If a user pastes /templates/{id}.json directly, fetch and validate the JSON document against the public schema.
Rendering the form
Render slots as editable text fields, look controls as optional choices or visual tiles, details as a freeform text area, and toggles as checkboxes or segmented controls. Do not hardcode Open Image Templates from this site; render from the schema so third-party templates work too.
Compiling the prompt
Treat user-edited slot values and selected controls as authoritative additions to the visible prompt. The final prompt should be inspectable before generation, and users should be able to copy it into another generator.
Model compatibility
The suggested_models field is a recommendation, not a lock-in. If your platform does not offer the suggested model, choose the closest compatible image model and keep the template prompt visible so users understand what changed.
Generation flow
A compatible platform can offer an Import Template button, a Use Template button, or a paste-URL flow. After generation, keep the template id, schema version, selected controls, generated date, and model metadata with the output so provenance stays intact.
Attribution and openness
Show creator and license metadata when available. If you publish generated examples or derivative templates, preserve the source template URL, schema version, and visible prompt. Do not market hidden prompt layers as Open Image Templates.
Minimum compatible implementation
At minimum, a platform should accept a template JSON URL, validate standard and schema_version, render slots, compile a visible prompt, generate with one image model, and show a link back to the original template.