I'm using GPT-3 in my bachelor's thesis and I need some assistance finding resources

I’m supposed to make a web app for kids that generates a fairy tale from pictures.

It’s supposed to go like this:

  1. Child chooses 1 (of 5) pictures
  2. Each picture is supposed to represent the start of a story with hard-coded text
  3. A request to the API is sent with the hard-coded prompt
  4. The text, along with the completion from the response are displayed
  5. Again, 5 pictures are shown and 1 is supposed to be picked.
  6. The text that corresponds to the picture is appended to the existing story and sent as a request

That’s basically it, I’d maybe need to add another set of images (so 15 total).

Everything is simple enough, but my main problem is finding the pictures. They have to be interesting and with enough details so that I can write a description of it and use it as a prompt. I searched on freepik and similar sites but the designs were too simple.
Any help at all is greatly appreciated.

I am not quite sure what exact type of pictures you are looking for, but may I suggest https://www.artbreeder.com/. It basically has a bunch of pre-trained GAN models (AI for training and generating images), some for very specific purposes (generating portraits or landscapes), some that are more general purpose (can generate anything from foods to vehicles to animals). Deep down it isn’t much different that the open-source GAN stuff that you can do on your own machine for free, but they wrapped it into a very nice interface, and have done all the drudgery of finding images to train the models, and so it is super easy to use.

Good luck.

1 Like