Pics: Interaction with the Environment is Key.
If you're rendering images, amongst some other layers that make all the difference, 1 - Plain prompt (no storytelling): "A dog running near a pond with a duck in the water." 2 - Storytelling prompt (with interaction and narrative): "A golden retriever excitedly chases a duck across the misty pond at dawn, water splashing as the early light filters through tall reeds, capturing the thrill of the hunt and the serene morning atmosphere." AND, if we then learn how to tell the AI to do an automated meta prompt, with all the roles needed etc, automatically we get to 3. THIS makes all the difference. Expand them into layers. P.S: 4 is a bonus when I told perplexity to use this meta prompt on its own, I just told the scene: A cat hunting a tiny spider, close-up.