Recipe Generation

There are various ways to generate recipes using the OpenAI API from the prompt wording and the parameters set. The API also has multiple engines to use but the most complex and the best engines to use are Davinci and Curie. Davinci is .. Talk about instruct beta as well


Due to an infinite way to generate recipes, I decided to use a more simplistic route by either only inputting a food name/type or a list of ingredients. While I could add more info and have the result more deterministic, I find that less info leads to more interesting results. Also,

Input = Name

Input = Ingredients

Introduction Generation

Name Generation

Input = Name

The Name i use are general types of food, such as burger.

GPT3 input is Write a <name> recipe: for Davinci instruct beta or <name> recipe:\nIngredients: for Davinci. I use the non instruct Davinci when I want to extend a recipe or make it more personal.

I then request GPT3 to come up with a name and introduction for the recipe:

Input = Write a name for this recipe: <recipe> in davinci instruct beta and Write an introduction for this recipe: <recipe>

Input = ingredients

I include ingredients and gpt3 makes instructions for it.