Inspiration
Millions of YouTube videos are uploaded daily, yet only a few go viral. We were inspired by the gap between effort and outcome—creators often guess what works. We wanted to turn virality into something data-driven and actionable.
What it does
Virality Coach predicts how likely a YouTube video is to trend based on its title, tags, and publish time. It generates optimised title suggestions and gives practical recommendations to improve performance instantly.
How we built it
We used a 2026 YouTube trending dataset across multiple countries. Using Python, Pandas, and Scikit-learn, we engineered features such as title length, word patterns, and publish timing. A Random Forest model was trained to classify viral content. We then built a Flask web app to make the model interactive and user-friendly.
Challenges we ran into
One major challenge was that raw model probabilities were too low and not user-friendly. We solved this by transforming the output into a more interpretable virality score. Another challenge was generating meaningful title suggestions that consistently improved scores rather than lowering them.
Accomplishments that we're proud of
We built a fully functional product, not just an analysis. The app provides real-time predictions, ranks multiple title options, and gives actionable feedback. We also successfully improved model performance and usability under tight time constraints.
What we learned
We learned how to turn raw data into a working product, not just insights. Feature engineering and model interpretation were key. We also learned the importance of UX—how data is presented matters just as much as the model itself.
What's next for Virality Coach
We plan to improve the model using NLP techniques for deeper text understanding, add thumbnail and engagement predictions, and personalise recommendations based on channel size and audience. Long-term, this could become a full creator optimisation platform.
Log in or sign up for Devpost to join the conversation.