The Blend

What it does

We've developed a radio app that listens to the user, understands their requirements, and with the help of NLP, Gemini LLMs, and a song recommendation algorithm, creates a personalized and tailored radio listening experience for them.

We used Flutter to develop the user-facing app, which communicates with a server developed using Express. This Express server is responsible for receiving the prompt, tokenizing it, generating a personalized blend, and delivering it to the user in a timely manner. The Express server communicates with another server developed using Flask. The Flask microservice handles data scraping, script generation for links using Gemini, and humanizing the script using Eleven Labs TTS. Both servers share a MongoDB database.

Gemini uses the scraped data and several other parameters, such as the time of day and mood, to generate a script for the link that should be spoken by the RJ. This script is then sent to the Eleven Labs TTS API to generate vocal data. The vocal data is processed to create a final link tailored specifically for the user.