Adaptive Music for Fitness and Wellness: Music That Follows Your Body

Most workout playlists are a guess. Someone picks songs that feel energetic, and your legs either match them or they don't. Adaptive fitness music turns that around: the music measures what your body is doing and moves to meet it.

This is one of the best-researched corners of adaptive music, partly because the effect is easy to measure. Here is what the studies found, what products have tried, and what it suggests for the next generation of workout music.

Why the beat matters: the synchronization research

Sport psychologists distinguish between asynchronous music, which plays in the background, and synchronous music, where your movements line up with the beat. A 2012 study in The Journal of Sports Medicine and Physical Fitness by Bacon, Myers and Karageorghis had ten untrained men cycle at 65 pedal revolutions per minute for 12 minutes while hearing music at 123, 130 or 137 beats per minute. Oxygen use was lower in the synchronous condition than in the slow asynchronous one (1.80 versus 1.94 liters per minute, roughly 7 percent).

The authors concluded that exercise is more efficient when performed in time with music than when the musical tempo is slightly slower than the movement. It was a small study, and the comparison was against slightly-too-slow music specifically. Still, it points at the core idea behind adaptive fitness music: the match between movement and beat does real work.

Approach one: choose the right song

The simplest adaptive system picks the next track based on your body. MPTrain, a research prototype by Nuria Oliver and Fernando Flores-Mangas of Microsoft Research presented at MobileHCI 2006, paired physiological sensors with a mobile phone. The user set a desired heart-rate pattern for the workout. The system monitored heart rate and step rate, then chose an unplayed song with a tempo close to the runner's current gait, nudged faster or slower depending on how far heart rate was from the target. The paper's early tests used a 30-song library ranging from 65 to 185 BPM and a single runner, so it was a proof of concept.

Spotify brought a consumer version to market in May 2015. Spotify Running used the phone's sensors to detect a runner's pace and picked music to match. According to a notice on Spotify's community forum, the company retired the Running feature on February 26, 2018.

Approach two: stretch the song to your stride

Choosing songs only gets you close, since your exact step rate rarely matches a song's tempo. The D-Jogger project at Ghent University went further. Body-worn sensors detected footfalls in real time, and a phase vocoder adjusted the music's tempo by up to about 10 percent in either direction without changing its pitch.

The 2014 PLOS ONE paper on D-Jogger, by Bart Moens, Marc Leman and colleagues, compared four strategies. Matching tempo alone produced modest synchronization. Starting each song in phase with the runner's steps did far better, and continuously adjusting both tempo and phase did best of all. One observation from the authors is useful for anyone designing workout music: once the lock between steps and beat was found, keeping it took less effort than finding it in the first place.

Approach three: produce the song for every tempo

Time-stretching has limits, and a song pushed too far starts to sound wrong. The third approach is to produce music designed to work across tempos from the start.

Spotify's 2015 launch included a set of original running tracks built this way. TechCrunch quoted Spotify's Gustav Söderström saying they were not simple beat stretching and that the composition itself seems to rearrange to fit your current pace. Weav, the company started by Google Maps co-founder Lars Rasmussen, built a running app around the same idea. TechCrunch reported in 2017 that producers used Weav's tools to shape how a song should change across tempos from 100 to 240 BPM, work Rasmussen compared to the last five percent of a remix, and that Weav had signed deals with Universal, Sony and Warner to build its catalog of adaptive tracks.

The takeaway is that wide tempo ranges are a production decision. Someone has to decide what a song should sound like at 140 and at 180, and that decision belongs to producers as much as to software.

Wellness: when the goal is to slow down

The same loop can run in the other direction. Wellness apps aim to bring heart rate and breathing down instead of pushing pace up, and the research on music and stress is substantial. We cover that side, including soundscape apps and the critiques of them, in humanistic adaptive music and music that knows the room.

The design questions are the same in both directions: which signal you follow, how quickly the music should respond, and how much control the listener keeps.

What this means for workout music next

Songs produced in several styles and playable at different tempos are natural raw material for this kind of experience. On Starchild, listeners can already change a song's tempo and key in the player and switch between full productions without losing their place. Browse by mood or genre and try one composition at a few different speeds to hear how it holds up.

• Beat alignment matters more than raw tempo. D-Jogger's best results came from matching phase, so the beat lands with the footfall.
• Small adjustments are easier to hide than large ones. Stretching a song a few percent is far less noticeable than doubling its speed.
• Wide ranges need authored versions. Weav and Spotify's originals both relied on producers deciding how a song changes at different speeds.
• Listener control still counts. Some runners want the music to lead; others want to set the pace themselves and have the music follow.

Sources

• Brunel University Research Archive: Effect of music-movement synchrony on exercise oxygen consumption (Bacon, Myers, Karageorghis, 2012)
• PLOS ONE via PMC: Encouraging Spontaneous Synchronisation with D-Jogger, an Adaptive Music Player That Aligns Movement and Music (2014)
• MobileHCI 2006: MPTrain, A Mobile, Music and Physiology-Based Personal Trainer (Oliver and Flores-Mangas)
• TechCrunch: Spotify for runners (May 20, 2015)
• Spotify Community: Retirement of our Running Feature
• TechCrunch: Google Maps co-founder Lars Rasmussen wants to make running fun through music (2017)