Understanding Audio Latency for Musicians: What’s Acceptable for Real-Time Online Collaboration?

When musicians are performing live, they need to hear each other in real time to maintain a solid tempo and rhythm. This makes maintaining low audio latency essential for online music collaboration as excessively delayed audio will quickly make a session unplayable.

But exactly how much audio latency is acceptable for rehearsing with other musicians online? Thankfully, there’s been enough research to provide a good idea of where the threshold lies. In this researched-backed rundown, we’ll discuss what audio latency is, what causes it, and what the data and our own research suggests is the upper limit for real-time music collaboration.

Quick facts

• Numerous studies have confirmed that music collaboration over internet networks can be effective and enjoyable.
  
  
• Studies indicate that effective online music collaboration requires latency levels of approximately 35-45 milliseconds (ms) or less.
  
  
• Data shows that increases in audio latency tend to result in players slowing down the tempo, but this issue can be solved by having a single player lead the group using a metronome.
  
  
• Most studies don’t account for latency jitter (changing latency levels), which is a challenge present on all internet networks. Latency jitter on internet networks has improved substantially in recent years, but more research is needed to learn how jitter impacts the upper limits of audio latency for musicians.

What is audio latency?

In a musical context, audio latency is the delay between when a note is played and when it is heard. Musicians are always subject to some degree of audio latency when performing live. Sound travels at roughly one foot per millisecond (ms), so even sitting three feet from an electric guitar amp, you would experience about 3ms of audio latency while playing the instrument.

With online rehearsing and jamming, typically the audio latency of your own instrument isn’t the issue, as it doesn’t need to travel over miles of fiber cable to reach your ears. But the incoming audio of your bandmates does need to go over the network before you can hear it, and if it is significantly delayed, it can make it difficult or even impossible to play in time. 

What causes audio latency?

When playing with musicians online, there are several sources of latency, with the most consequential being network latency:

1. Analog to digital audio conversion
   This is the process of turning an audio signal into digital information that can be processed by a computer and transmitted over the internet.
   
   
2.  Software latency (DAW)
   Software latency results from the time it takes audio software (usually a DAW) to process a signal and is often greatly increased by adding software effects and audio processing programs (plugins) to a digital audio signal.
   
   
3. Network latency
   Network Latency is the primary cause of latency for online music collaboration and results from the time it takes a digital audio signal to travel from one location to another over an internet network.
   
   
4. Digital to analog audio conversion
   Once the audio reaches its destination (your remote collaborators), it must be converted back into an analog signal before it can be played out of a collaborator’s headphones or speakers.
   

How much audio latency is acceptable for music collaboration?

What amount of audio latency is noticeable? Recent research indicates humans are capable of detecting audio delays shorter than a single millisecond, but that’s not really the number we’re after. The real question is, at what point does audio latency become a problem for musicians?

Most of the data suggest that approximately 35-45ms of audio latency is the maximum amount for a good online music collaboration experience, with sessions becoming unplayable once 45-55ms is reached. If one musician is designated as the tempo “leader,” and keeps tempo using a metronome, longer latency delays may be playable.

Synchronized Clapping Study

A 2002 study at Stanford tested subjects’ ability to clap a short repeating rhythm in sync with audio latency applied to their incoming audio. The subjects in this study were not required to be musicians. This study's author asserted that audio latency beyond 30ms made real-time collaboration impossible, a conclusion that mainly relied on data that showed the tempo consistently slowed down once latency levels went beyond 30ms. 

Interestingly, despite the EPT, the study also showed that subjects were able to stay in sync with up to 40ms of latency, and even up to 70ms of latency by using a leader/follower technique, in which one subject ignored their testing partner’s clapping and kept tempo on their own, with the “follower” adjusting to the leader’s tempo.
The 2002 study at Stanford provided evidence that most people should be able to maintain a synchronized rhythm at latency levels of 30-40ms and below, even if that rhythm was very likely to slow down past 30ms. However, there are considerations to keep in mind when applying this data to online music collaboration:

• Clapping a short, unchanging rhythm is nowhere near as challenging as performing a complex piece of music, but perhaps that balances out with the fact that these test subjects were most likely not practiced musicians.
  
  
• The Stanford study did not take place over an internet network, but used a computer program to manufacture precise audio delays. A real internet network introduces inconsistencies, like latency jitter, adding more complexity and challenges.
  
  
• As we’re about to cover, later similar studies did corroborate the slowing of tempo with increased latency, but did not support the impossibility of music collaboration beyond 30ms of latency. 
  
  

Simulated remote performances with music students

A 2006 Study by the Eastman School of Music found that music collaboration was not only possible, but enjoyable beyond the 30ms audio latency limit asserted by the Stanford study. The 2006 Eastman study tested two pairs of musicians’ ability to perform a classical piece with different levels of audio latency applied to their incoming audio. The test subjects were undergrad students at the Eastman School of Music with approximately 10 years of music study. One pair played wind instruments and the other strings. The students were given music scores they didn’t know, but relatively easy pieces were chosen: Mozart’s K. 487, Nos. 2 & No. 5, duets. The music students rated the “musicality” of each performance.

The musicians reported being satisfied with latency levels as high as 46ms and gave favorable ratings for latency levels up to 56ms. The musicians experienced substantial difficulty at 86ms of latency, and the performances broke down at 126ms of latency. Like the Stanford hand-clapping experiments, this study found a fairly consistent pattern of tempo slowdown with increased latency, though with some exceptions.

Again, there are some details to keep in mind before applying these findings to online music collaboration:

• Like the 2002 Stanford study, the 2006 Eastman study did not test using actual network latency, but precise and consistent manufactured audio delays that did not account for the inconsistencies, like latency jitter, that one would experience on any online platform.
  
  
• The instruments chosen (wind and strings) are both capable of slow attack transients, which may have been more forgiving in high-latency environments, depending on how the musicians interpreted the assigned pieces.
  
  
• One of the biggest takeaways from the Eastman study is that music collaboration is possible at latency levels beyond 30ms.
  
  

Recent simulated remote performances

A 2025 Study published in Future Internet tested 11 pairs of musicians with a diverse range of instruments and styles, including rock and pop arrangements. Musicians chose their own music, performed in isolated rooms, and could see a video feed of their testing partner. Precise analog delays were used to introduce audio latency levels from 0 to 60ms. Questionnaires and, to a lesser extent, videos were used to gauge the quality of sessions.

 The 2025 Future Internet study found that musicians could stay in sync and have high-quality collaboration experiences with audio latency levels up to 40ms. It should be noted that there was not a steep drop-off in quality after 40ms, but a gradual decline in reported quality and satisfaction. At 60ms of audio latency, musicians were not able to reliably stay synced. This study also found that tempos began to slow down with increased audio latency. Like the previous studies, this experiment also did not account for latency jitter.

Lutefish audio latency research

At Lutefish, we also did our own research to determine acceptable latency levels for online music collaboration. Pairs of musicians played together in separate sound isolation booths, with different amounts of latency applied to their incoming audio. Musicians rated the experience at each level of latency. Musicians rated their experiences as “good” or better with latency levels up to 40ms and playable with latency levels up to 60ms. On average, detrimental effects began at 32.5ms, and sessions became unenjoyable at 55ms.

What the data says about latency and live music collaboration

If we average the upper limits of acceptable audio latency across the above studies, we get an approximate acceptable latency of 41.5 milliseconds. That’s about the level of audio latency you experience every time you hear a sound that originates 41 feet away from you. Many of these studies increased latency levels in large degrees rather than gradual slopes, and the studies that did record gradual changes in audio latency did not record steep drop-offs in quality or performance one latency increased above 40ms. With that in mind, it’s unlikely that 41.5 milliseconds is a hard limit, but more of a general indicator.

Other factors may also have a significant impact on level of acceptable audio latency for a performance, including:

• Style of music
  
  
• Type of instrument
  
  
• Player experience
  
  

When considering the available data, a range of 35-45ms is likely a more useful indicator than the 41.5ms average when determining the maximum amount of latency for real-time online music collaboration.