Mono to Stereo: How AI Is Breathing New Life into Music

Here is the summary of the article in 10 sentences: Music in mono format lacks the spatial richness that makes it feel alive. With AI, it's possible to transform mono recordings into stereo or remix existing stereo recordings. The use cases for mono-to-stereo upmixing include enriching existing mono music to stereo and improving or modernizing existing stereo mixes. For instance, many popular songs from the 1950s and 1960s were recorded in mono, such as those by Elvis Presley, Chuck Berry, and Duke Ellington. AI can convert these recordings into stereo, making them sound more lively and engaging. The process involves different techniques, including traditional signal processing, source separation, machine learning with parametric stereo, and generative AI. For example, a 2007 paper proposed an algorithm that identifies different sound sources in a mono recording, extracts them, and then mixes them back together for a realistic stereo experience. More recent advancements in AI have made it possible to use stem splitting tools or machine learning models to predict spatial parameters from a mono signal and create a stereo experience. However, the field of mono-to-stereo upmixing research is sparse and requires more openly available demos and code to progress.