This workflow demonstrates using AI Agents as experts in their respective fields to create better image generation prompts. It uses Griptape's ComfyUI nodes (https://github.com/griptape-ai/ComfyUI-Griptape), OpenAi's gpt-4o, dall-e-3, and ollama with llama3.1.

In the example image provided, the prompt was "stuffed bunny". The agents provided additional context automatically, giving feedback on color, cinematography, detail, and used websearches to provide proper prompt information.

Based on their feedback, "stuffed bunny" became:

"Stuffed bunny, medium close-up, positioned on a plush cushion in a cozy children's bedroom at dusk. Captured with a Canon EOS 5D Mark IV, 50mm lens. The scene is bathed in soft golden light, creating a warm glow with muted earthy tones of beige, cream, and soft gray. The bunny has pastel pink fur, centered in the frame, slightly above eye level. It features a faded plush toy appearance, with a worn-out cotton ball tail, lopsided head, matted and frayed fur, glassy plastic eyes, and a hand-sewn patch on its left ear."