AI Helps Musician Restore Vintage Irish Fiddle Music
Michael Coleman’s recordings shaped Irish traditional music throughout the 20th century, but early recording technology left these historic performances locked in challenging audio formats–world-champion Irish accompanist Colman Connolly set out to change that.
“I grew up in an Irish American musical family, and I was surrounded by traditional Irish music from a young age,” Connolly explains. “Michael Coleman’s name was a constant presence… I felt a strong connection to his legacy.” Connolly’s project aimed to restore and remix Coleman’s fragile 78 rpm recordings, many of which suffer from hiss, pitch wobble, and tightly blended mono tracks.
“The main goal was to restore both the sound quality and the cultural authenticity of his work,” Connolly says. “I thought it would be cool to use AI to bridge the past and present while honoring my heritage.”
The technical challenges were significant. Early on, Connolly ran into difficulty isolating the fiddle from the piano accompaniment. “I tried tools like Moises and Lalala AI to isolate the stems,” he recalls, “but with some, the results were a complete blank, which was frustrating. Others showed promise with guitar tracks but struggled with bleed between piano and fiddle stems.”
That’s when his thesis advisor, Daniel Rowland, introduced him to AudioShake.
Thanks to AudioShake’s industry-leading AI stem separation technology—featuring specialized string separation models and best-in-class audio processing—Connolly was able to isolate clean, professional-grade string, piano, and guitar stems from century-old mono recordings.
“AudioShake was able to take those original tracks from the 1920s and 1930s and spit out isolated fiddle and isolate the piano,” Connolly says. “It was a game changer. What I thought was at first an impossible, challenging project became one that was feasible.”
Using the isolated stems, Connolly restored the clarity of Coleman’s original fiddle recordings and composed new piano parts that aligned harmonically and rhythmically with the fiddle, being careful to adjust the piano tone to fit the era and blend seamlessly with the originals.
The thesis resulted in 22 remixes of the original and 15 modern takes on Coleman’s original performances paired with Connolly’s piano. “It was a labor of love,” Connolly reflects, “blending technology and AI with my passion for Irish music.”
By bridging historic recordings with modern technology, Colman Connolly’s work highlights how tools like AudioShake can preserve cultural heritage while opening new possibilities for musicians and engineers worldwide. As archival audio continues to present unique challenges, AI-driven solutions are proving essential in unlocking and reimagining sounds that might otherwise remain lost to time.