Read & Accept EULA
Check if a Song is AI-generated - Beta

AI Music Checker

Check if a Song is fully or in parts AI-generated.
Please be invised this is an early beta version which we are training and improving regularly.
We cannot guarantee 100% accuracy just yet... But we are getting there. Enjoy! 🤓


Analyze Song AI-generation


OR


About AI Music Checker

Welcome to our AI Music Checker! We're excited to have you explore our innovative solution for identifying AI-generated songs. Here's a bit more about our work and how our model functions:


Our Work

The AI Music Checker is the result of extensive research and development, combining advanced machine learning with deep insights into musical characteristics. Our goal is to create a robust and reliable model that can identify whether a song is entirely or partially generated by AI.

We have been diligently working to gather a variety of songs, both AI-generated and human-created, to train our model. By analyzing all kinds of songs and music, we have developed a model capable of discerning subtle differences between AI-generated and traditional music.


Audio Features We Measure

Our model analyzes a range of audio features to determine if a song is AI-generated. Some of the key features we measure include:

  • MFCCs (Mel-Frequency Cepstral Coefficients): These coefficients represent the spectral content of sound and are crucial for identifying pitch and timbre.
  • Chroma Features: These features represent the twelve pitch classes of Western music and help us understand the harmonic structure of a song.
  • Spectral Contrast: This feature measures the difference in amplitude between different frequency bands, providing insight into the clarity and definition of the sound.
  • Pitch and Rhythm: We analyze pitch contours and rhythmic patterns to identify characteristic traits in AI-generated music.
  • Spectral Centroid and Bandwidth: These measurements give us information about the "center of gravity" and frequency spread of the sound, which can differ in AI-generated tracks.

Training Data

So far, we have trained our model on 128 songs. While this is just the beginning of our development and training for this model, it provides a solid foundation for our ongoing efforts. Our dataset includes a variety of genres and styles to ensure that our model is robust and can handle different types of music.


The Future

We are continuously working to improve our model. We regularly add new songs to our training dataset and adjust our algorithm to enhance accuracy and reliability. We are excited to see how the AI Music Checker will evolve and hope it will become a valuable resource for musicians, producers, and music enthusiasts worldwide.

Thank you for using AI Music Checker! We hope our model helps you better understand and appreciate the music you listen to.

/Joel