prompt-extend VS RecipeGPT-exp

I've been playing around with DALL-E 3 a lot recently. One of the things they do is to expand a user's prompt in order to add a lot of detail. Via their API, you can see the expanded prompt (whereas, you can't through the ChatGPT interface).

They obviously have the power of their LLM behind them and can generate some really interesting prompts. There is an open source implementation that the creator of Fooocus made which attempts to expand on prompts using some commonly used keywords[1] with some sort of basic context.

e.g., I typed in "Brisket on a table" and got: "Brisket on a table, product photography, Michelin star, award winning photo, 8k, trending, HD. High quality image, highly detailed, stunning lighting, flawless render, masterpiece, still from the movie directed by Denis Villeneuve with art direction"

You get a much better image with that prompt vs just the basic: "Brisket on a table"

[1] https://huggingface.co/spaces/daspartho/prompt-extend

🆓  Extend Prompts AI art is so cool, but I’ve found that to make eye-catching designs that resemble what you’ve imagined, you need to add a lot of prompts. This tool expands on the prompts you give it. You then enter these longer prompts into the AI design tool of your choice. The results are almost always more beautiful.

I used prompt-extender just for the sake of having something super random to work with. https://huggingface.co/spaces/daspartho/prompt-extend

There is already something similar called promptmaker, which a member of our discord server made.

Hey! The model files are available on HuggingFace Hub here You could use it directly using the HuggingFace library, check this notebook for reference

Here's the GitHub repo for the project, which contains all the code for the project. I've also uploaded the model and the tokenizer on HuggingFace Hub.

Great question. Models for generating plausible recipes have existed in some form for about a decade now.

Arguably the first "successful" attempt at this was Chef Watson, which blew my mind when it was first released in 2014 despite it's well-documented tendency to suggest all kinds of spectacularly odd combinations of flavors and ingredients, like garlic ice cream and mayonnaise-spiked Bloody Marys[1].

It's worth noting that preprocessing the textual inputs isn't entirely necessary to produce somewhat reasonable, ML-generated recipes. For example GPT-3 is capable of generating fairly interesting zero-shot recipes, despite having been trained on raw text data without any preliminary feature selection to label (e.g.) a recipe's ingredients.[2] Still not exempt from the occasional wacky, whimsical suggestion[3], but I, for one, wouldn't want my ML-generated recipes any other way.

1. https://www.google.com/amp/s/www.newyorker.com/magazine/2016...

2. https://github.com/LARC-CMU-SMU/RecipeGPT-exp

3. https://thenextweb.com/news/ai-generated-recipes-three-cours...