Similólúwa Òkunọwọ
Back to projects
3 min read
#Audio Analytics#Streamlit#AssemblyAI#LangChain

MusicCritique

MusicCritique leverages audio transcription and language models to deliver actionable insights that help artists refine their songs and enable listeners to better understand the music.

Live Demo GitHub

Streamlit LLM Hackathon 2023 Winner 🎉

Introduction

MusicCritique is an application that harnesses the power of audio transcription and LLMs to provide artists/music creators with actionable insights from their recorded songs. It empowers music creators to make more informed and strategic decisions before releasing their musical masterpieces to the world.

MusicCritique also enables listeners to better understand music and make listening decisions through audio analytics.

Key Features

  1. Audio Transcription: Designed a pipeline using AssemblyAI APIs for converting audio and video recordings into text for easy analysis.
  2. Summarization: Provides a concise summary of the song’s content.
  3. Sentiment Analysis: Analyzes the emotional tone of the song to gauge its mood.
  4. Lyrical Content Analysis: Identifies key themes and keywords in the song’s lyrics using visualizations.
  5. Content Moderation: Detects sensitive or inappropriate content in the song’s lyrics.
  6. Topic Detection: Identifies the main topics discussed in the song.
  7. Criticism: Offers a constructive critical analysis based on various song elements.
  8. Recommendations: Provides actionable suggestions for improving the song based on the analysis.

Technologies Used

  • Python
  • Streamlit: For the UI
  • Assembly AI: For audio transcription and PII content extraction
  • LangChain: For generating critical analyses and recommendations using OpenAI’s GPT 3
  • StableDiffusion: For generating cover image ideas
  • Plotly: For visualizations
  • Pytest and SeleniumBase: For testing

Potential Use-Cases

  • Providing independent artists with actionable data-driven insights to refine their songs by analyzing lyrics, sentiment, and emotional tone before release.
  • Helping music managers to ensure that their artists’ songs align with the artist’s brand and audience through content analysis and feedback.
  • Helping record labels evaluate songs’ potential for market success and detect sensitive content before release.
  • Providing music listeners and critics with detailed, data-driven reviews based on lyrical and sentiment analysis.
  • Matching music with emotional tones and topics that resonate with target audiences for campaigns.