How to Look for Plagiarism in (AI) Music

How to Look for Plagiarism in (AI) Music - Complete Guide

I was experimenting with the beta version of Yudo at yo.com, trying to generate an Indie Folk song. After several iterations, the music generator produced a tune I enjoyed and thought could potentially be published as music. I then wanted to expand on this idea by creating a music video using available AI tools for image and video generation. As a result, I pieced together a 100% AI-generated music video.

I plan to further develop this project and share the workflows and steps involved. In the meantime, you can view the entire music video and listen to the song titled "Dances of Daylight" by skipping to the last chapter or timestamp in the description below.

According to Yudo's terms of service, full rights to the music generated are granted, but they do not take responsibility for any copyright claims that may arise. This is fair enough, considering even traditional musicians can face unintended similarities to existing works and therefore risk copyright infringement. The same risk applies to AI-generated music. Developers of AI music generators should ideally train their models to avoid plagiarism, but assumptions need verification.

To verify if the music I created has potential copyright issues, I discovered three types of plagiarism detectors:

Detectors analyzing the musical composition.
Detectors analyzing the lyrics.
Recognition applications for plagiarism by exclusion.

Musical Composition Analysis

There is one tool that accomplishes musical composition analysis—MIP (mip.ai.com), a newly launched website based in Korea. Users can submit songs for analysis, and the software then generates reports on potential plagiarism by comparing songs to a database of thousands of tracks. The process involves breaking down and analyzing song structure, rhythm, melody, and harmony.

Testing my AI-generated song "Dances of Daylight" returned a 22% similarity rate to "Blow Me One Last Kiss" by Pink, which is within the acceptable range. However, the results for Mumford & Sons' "Winter Winds" were mixed, highlighting potential database limitations.

Lyrics Analysis

For lyric analysis, tools like Grammarly's Plagiarism Checker and Quetext can be used. Both are straightforward and effective at identifying potential plagiarism in song lyrics. In my case, the analysis showed no significant plagiarism risk, and even mentioned the Bible as a reference, which isn't something to worry about.

Song Recognition

Song recognition apps like Shazam and Google's music recognizer can help detect potential plagiarism by exclusion. Shazam requires exact matches, while Google's recognizer offers more flexibility by allowing melodies to be hummed or sung.

Web-based apps like AHA Music and ACRCloud also offer sophisticated recognition capabilities. ACRCloud even provides detailed reports by analyzing uploaded files. Although these recognizers have their limitations, current advancements in AI and computing make future improvements highly likely.

Uploading music to platforms like YouTube or Spotify provides an additional method of copyright infringement detection due to their extensive and constantly updated music databases.

Conclusion

While it seems my AI-generated song is clear of major plagiarism concerns, inherent risks remain for both AI and traditional music creators. Vigilance and the use of multiple detection tools can help mitigate these risks.

Keywords

AI-generated music
Plagiarism detection
Musical composition analysis
Lyric analysis
Song recognition apps
Yudo music generator
Copyright infringement

FAQ

Q: Is AI-generated music at risk of plagiarism? A: Yes, just like traditionally composed music, AI-generated music can unintentionally resemble existing works and risk plagiarism.

Q: What tools can be used to check for plagiarism in AI-generated music? A: Tools include musical composition analysis (MIP), lyric analysis (Grammarly, Quetext), and song recognition apps (Shazam, Google Music Recognizer, AHA Music, ACRCloud).

Q: How reliable are current AI music plagiarism detectors? A: While promising, current detectors are still evolving, and some limitations exist, particularly with genre coverage and database comprehensiveness.

Q: Can I upload my AI-generated music to platforms like YouTube without worrying about copyright issues? A: Uploading your music can help as these platforms have extensive databases and automated copyright checks. However, there is always a risk, and it's best to use multiple detection tools for thorough analysis.

Q: What should I do if my AI-generated song is flagged for plagiarism? A: If flagged, consult with legal experts to understand the claim's validity and potentially make necessary adjustments to your music.