Inspiration
As content creators music for a tik tok video serves as the core reason of getting attention from our followers and good feedback. Using music curated to the emotional quotient of the video was our main challenge. With vast options to choose from, finding the best fit is time difficult especially the non trending ones.
What it does
Our application uses SOTA AI algorithms to understand the mood of the video and suggest music that suits it. It factors in both vocals and instrumental part of the music we are trying to recommend. This gives a mathematically calculated music for tik tok videos.
How we built it
We used an AI algorithm to extract emotion and encode them to do a similarity search on the database containing music records.
Challenges we ran into
Latency was one of our main challenges.
Accomplishments that we're proud of
We used pinecone as a vector database to calculate similarity of emotions using 9 dimensions. Collaborating with a team. Training an auto encode from scratch
What we learned
using vector databases and its applications.
What's next for Ongaku
Integrating Onkagu into the current tik tok's music discovery mechanism by actively participating in their teams full time.
Log in or sign up for Devpost to join the conversation.