a little project that allows users to extract recipes from social media videos (TT, IG and YT) using AI (Gemini Pro 2.5).
uses react + postgres, mainly built using agent V2. found that agent V2 is much improved and pretty solid. used assistant much more previously but now it seems the other way around.
tried to extract thumbnails from the videos but unfortunately IG and TT have protections against that.
This is nice, Next to copy button could a “Print This” button be added? That makes a print friendly version of the recipe.
Another idea I have is for adding a unique QR code to the corner of each printed page so that someone can just scan the code and order the food items on amazon effortlessly.
I made a shopping app back in 2018 and wish Insta-cart had a api you could utilize they would get so much more business.
Tried testing with YT shorts and long: both said “only YT, IG and TT supported” or similar. Keen to know how you ‘read’ a video. Look for on-screen text? Voice to text? And how are you ingesting vids from these platforms? Sounds fascinating.
thanks mate, i will add a print button shortly. and the unique QR code sounds like a great idea - will definitely look into this. do you have any experience with adding QR codes?
hi chris, what URLs did you try to submit? i have not had any issues myself so i am curious.
as for the reading of videos, you can ask Gemini Pro yourself for more details but In essence, Gemini Pro acts like a highly attentive and knowledgeable assistant who watches the video, listens carefully, reads any text, understands the sequence of actions, identifies all the key components of a recipe, and then neatly writes it all down in a standard format.
i also have some custom logic that helps the AI in getting it right.