Custom Gems
Custom Gems is a technique for controlling image generation in ChatGPT Image 2.0 through structured JSON-formatted prompts processed via the Gemini 2.5 Flash API. Rather than relying solely on natural language instructions, this approach enables users to specify parameters and requirements in a machine-readable format, providing granular control over the output generation process.
Implementation
The method works by encoding image generation specifications into JSON structures that are then passed to Gemini 2.5 Flash for processing. This intermediary step allows for more precise parameter definition and constraint specification than conventional text prompts alone. The JSON format standardizes how requirements are communicated to the image generation system, reducing ambiguity in instruction interpretation.
Use Cases
Custom Gems are particularly useful in workflows requiring consistent, repeatable image generation with specific technical or aesthetic requirements. By defining parameters explicitly in machine-readable form, users can achieve more predictable results across multiple generation attempts and more easily automate image creation pipelines within larger AI agent systems.