The musician uploads an audio file through the React interface. Firebase supports the storage and database layer.
What I built
A React and TypeScript frontend connected to Firebase storage and database services, with Cloud Run inference for audio extraction and analysis.
I designed the end-to-end extraction and inference pipeline around an upload-only workflow, returning ML-driven originality and quality signals.
Purpose & audience
Help artists and teams verify song originality before release using automated music analysis.
Built for musicians, producers, and music teams evaluating originality, quality, and marketability.
Under the hood.
Follow the architecture, one connection at a time.
SongSearch / architecture
The pipeline extracts audio features for analysis, requiring no additional user input beyond the uploaded file.
Cloud Run inference services combine lyrical identification, motif repetition, rhythmic predictability, harmonic tension, and information-theoretic metrics.
The upload-only workflow returns ML-driven originality and quality signals for musicians and music teams to consider before release.
Scroll to follow the connections. Select a step to take a closer look.
Decisions & tradeoffs
I combined information-theoretic metrics with lyrical identification, motif repetition, rhythmic predictability, and harmonic tension.
Keeping file upload as the only required input makes the tool approachable. Bringing several musical signals together provides more context than relying on one measure of originality.
Outcomes & limitations
Built an end-to-end cloud music-analysis workflow and documented its architecture in the public case study.
Originality and quality still require musical judgment. The analysis provides signals rather than a definitive creative verdict.
What’s next
Possible directions for further development.
Explore how the analysis compares with feedback from musicians and producers, while keeping the individual signals understandable. Creative judgment should stay part of the workflow.