💬 Uncle Dave says...
I decided to use this recipe to compare AI image creation. These are, in order:
- turk_veg_cop_9778 - CoPilot
- turk_veg_gpt_31730 - ChatGPT
- turk_veg_meta_31035 - Meta AI (I chose the best one of four it created)
- turk_veg_grok_47055 - X Grok (I chose the best of two it created)
- turk_veg_gem_23413 - Google Gemini
- turk-veg-aria_13965 - Aria, which is free with Opera Browser.
😞I feel like the Copilot and ChatGPT ones were the least realistic.
🎯And as is often the case, the one from Aria stood out as one of the best.
🪄The image creation prompt was created by Claude AI (which does not itself make images) based on recipe information I gave it, including name, ingredients, directions, the "blurb" that tells about the recipe, and any comments that are in Uncle Dave Says... It has basic standing instructions** in its project file for what to do when I ask it to write an image creation prompt for me.
✏️This is the prompt it created:
"Please create an image of a steaming bowl of creamy turkey-veggie soup. The soup should have a rich, velvety texture with visible chunks of roasted turkey meat, orange sweet potatoes, green zucchini and broccoli florets. The soup should be served in an elegant, white ceramic bowl on a rustic wooden table with a cloth napkin and silver spoon nearby. Include a small basket of crusty artisan bread in the background. The lighting should be warm and inviting, suggesting comfort food on a cool day. The image should have a shallow depth of field with the bowl in sharp focus, styled like professional food photography with natural-looking steam rising from the soup. The colors should be vibrant yet warm, showcasing the fresh vegetables and creamy broth."
** The standing instructions Claude has for creating images:
'When I provide you with the recipe information, create a prompt that will allow image creating AIs to create an enticing and professional looking image of the resulting dish.
Make your prompts less than 1024 characters whenever possible.
I may include additional instructions, which will appear inside [...] at the top of the text.
Begin the prompts with "Please create an image..."'