I have always been a foodie & culinary enthusiast. I love to cook and develop recipes, a hobby that took root back in my university days. Over the years, I have accumulated numerous photos of dishes I have made and developed, but unfortunately, I have misplaced a lot of the recipes.
Now, with my interest and enthusiasm leaning towards AI, I decided to use DALL-E & GPT Vision to breathe new life into my old recipes. In this blog, I will walk you through how I designed the prompt for each photo and the task given to rejuvenate the photos with AI, in hopes of compiling them into a cookbook.
Prompt Design
The tasks I needed to highlight in this prompt design involved
- Uploading an image of a dish. I need to instruct GPT that there was an image in GPT Vision for it to use in the tasks.
- Providing a detailed description of the dish in the image, including: The main ingredients featured in the dish. Cooking methods used (e.g., sautéed, baked, grilled). Any notable textures or colors in the dish. The type of garnish or additional elements for visual appeal.
- The next part of the prompt should clearly ask for a recipe to be generated with ingredients & specific quantities. Step-by-step instructions for cooking and any particular cooking techniques or tips related to the dish.
- Finally, request the generation of two images that: Offer different angles of the dish to provide a comprehensive view. Maintaining a cohesive theme, such as focusing on the dish’s aesthetic or highlighting specific ingredients.
To illustrate this prompt design, here’s an example of one of my recipes: “Pan-Sautéed Diced Turkey Breast in Cranberry Sauce with Spring Onions.”
I provided ChatGPT-4 with the ingredients and the cooking method, along with the desired flavor profile and presentation. The AI then weaved this information into a coherent, easy-to-follow recipe, complete with culinary techniques and garnishing tips.
My prompts were meticulously designed to ensure that the resulting text and images were not only accurate but also evoked the right image resembling the original dish. As you can see below, DALL-E generate 2 images based on the uploaded dish I provided and I was pretty happy with the results.
I then went on to test my prompt on another dish Pan-Fried Chicken Breast with Thyme and Beetroot Infused Basmati Rice:
Images of the dish that DALL-E generated for Pan-Fried Chicken Breast with Thyme and Beetroot Infused Basmati Rice.
In creating this prompt, I realised that the more descriptive I am of the uploaded dish the better the image generated by DALL-E.Overall the goal was to showcase the potential of AI in creative endeavors, it is a testament to how technology can be harnessed to enhance our creativity. Eventually I plan to generate enough recipes to create a AI-generated cookbook based on my collection of recipes.
Leave a Reply