Inspiration
We wanted to do something unique for the crystal ball challenge, and we thought that an algorithm which could predict how many views a TikTok video will get would be an interesting project.
What it does
The user is able to interact with a basic chatbot and must answer a series of prompts (e.g. their number of followers and account name) before uploading their video. This is all done through a messenger-style interface.
Once they have entered the relevant data and uploaded their video, the algorithm will analyse random frames in their video to generate a description of it. This includes any hashtags it thinks could be included in the video description. This description, along with the length of the video, the number of followers, etc. will all be compared against a statistical model generated from 100s of TikTok videos and an estimate of the number of views the video will achieve will be calculated and shared with the user.
How we built it
We used a linear regression algorithm to determine how different factors, such as video length, etc. affect the number of views a TikTok will receive. We used TalkJS and some basic HTML, CSS, and JS to build the website along with Python to analyse the video.
Challenges we ran into
Unfamiliar with TalkJS, issues with importing certain modules in Python.
Accomplishments that we're proud of
We think that it is a fairly unique idea and has been implemented.
What we learned
How to use TalkJS to a decent level; gained experience with various AI-driven Python modules.
What's next for Crystal
Crystal's accuracy could greatly improved with a larger dataset. In addition to this, various UI changes could make it easier to interact with and some code optimisations would make it faster.
Log in or sign up for Devpost to join the conversation.