videoGPT

hey, as someone who consumes a ton of YouTube videos, i've always wanted something like videoGPT. when chatGPT with vision launched, i decided it's the right time. you can ask videoGPT questions about the video, and it will answer them while referring to the sources it used in the video. videoGPT will summarise the video in a paragraph by default as well there's a lot of apps & chrome extensions with a similar thing happening, but afaik this is the first and only to use GPT4-V to actually process the visual content of the video too. all others just analyse the subtitles / captions / descriptions. because of this, videoGPT does need some time to process to develop it's deeper understanding of the video there's a chrome extension coming soon. it's not mobile-friendly so use it on desktop for now please it's free but i'll be making a Pro plan soon, which will allow videos longer than 30 mins to be used, and will allow videos without subtitles to be processed it's powered by chatGPT, GPT-4V, CLIP & langchain /// coming soon /// chrome extension pro plan (with longer videos, and audio transcription) upload your own videos thanks! :) P.S: for devs ;0... the core aspect of videoGPT (and the code i'm most proud of) is the algorithm that understands the visual content of the video. doing this in a compute & cost efficient way is paramount, so i built a home-baked algorithm (in 800 lines of python ;) to collapse the video into a subset of key frames. these frames are then sent to GPT-4V to create a visual understanding of the video too.

Awesome! I will give it a try with some podcasts 😄 Congrats on the launch! 👏

Wow, videoGPT sounds like a game-changer! I'm curious to know how accurate the chatbot is in analyzing the key scenes of a video. Have you done any testing or received feedback from users? I think this could be incredibly useful for content creators and researchers. Looking forward to seeing the pro plan and exploring more features!