• Google Magenta developed a way to jam with a machine using different instruments
  • We wanted to make that possible for singers too

What it does

  • RAPPLE is your smart rap battle opponent
  • It listens to your freestyle on a beat
  • It then responds to you with a punch line on the same beat
  • It helps freestyle rappers to practice
  • It inspires songwriters in the creative process of texting

How we built it

  • RAPPLE is a modular system consisting of the following building blocks:
  • Speech Recognition by Microsoft Cognitive Services (transforms input rap audio signal to text)
  • Punch line generator (generates an appropriate response and outputs it as text)
  • Text-to-Speech by Microsoft Cognitive Services (transforms the text to a speech signal)
  • Speech-to-Rap (imposes rhythmic structure on the speech signal to make it groovy and fit it on the beat)
  • Output the punch line(s) as audio

Challenges we ran into

  • How to connect all the different parts
  • How to transform speech to rap
  • How to develop a good flow of the system
  • How to play the beat and the response simultaneously

Accomplishments that we're proud of

  • Some actually meaningful punchlines
  • We created a human - AI interaction
  • There is a lot of potential in this idea

What we learned

  • Everything takes time
  • How to make different software pieces play together smoothly

What's next for Rapple

  • Thanks to its modular structure it can be easily expanded to many use cases, including:
  • Adapt the musical style of the system's response to the style of the user input (Rap/Soul/Rock/Metal)
  • Singing duet partner: practice singing in two voices
  • Voice timbre matching: make RAPPLE reply with the voice of your favorite rapper
  • Snapchat/Instagram integration: Post a story of you battling your favorite rapper

Built With

Share this project: