PortAudio M1 Latency

October 4th, 2023
music, tech
I recently tried to get my whistle-controlled bass synthesizer working on my Mac, as a backup to my Raspberry Pi version. [1] It's written on top of PortAudio, a cross-platform audio library, so getting it to compile and run was mostly a matter of getting the library installed, which went quickly. Unfortunately, the latency was far too high to be useful as a real-time musical instrument.

Testing with PortAudio's demo paex_read_write_wire.c [2] with suggestedLatency changed from defaultHighOutputLatency to defaultLowInputLatency I see:

Input device # 2.
    Name: MacBook Pro Microphone
      LL: 0.0528542 s
      HL: 0.0621875 s
Output device # 3.
   Name: MacBook Pro Speakers
     LL: 0.0187083 s
     HL: 0.0280417 s
Num channels = 1.

This is 72ms of round-trip latency, or 3,156 samples. You can easily hear the latency, and it's pretty painful for live music. It's even higher than the 67ms I see in the web browser. [3] Changing FRAMES_PER_BUFFER from the default of 512 to 128 doesn't have any effect, so I think this is lower level buffering.

I also tried with a Scarlett 2i2 audio interface, and latency wasn't any lower. Setting PA_MIN_LATENCY_MSEC also doesn't do anything. I'd try turning off ambient noise reduction but this option doesn't exist for M1 Macs.

I don't have this problem when running the Reaper DAW so it's definitely possible to get low latency on this machine. I wonder if there's some kind of default processing running on the audio system, and Portaudio doesn't ask to turn it off?

(I could go back to running my code as a Reaper plug-in, but I'd much rather have something stand-alone.)


[1] In setting up at Adirondack Dance Weekend I couldn't find one of my two Pis. It turns out I'd missed that it was under my mixer and forgotten to pack it:

[2] Built with gcc -I/opt/homebrew/include/ -L/opt/homebrew/lib/ -lportaudio paex_read_write_wire.c -o paex_read_write_wire after brew install portaudio to get the library.

[3] Suspiciously this is quite a bit higher than I used to see on my Intel Mac. I used to get 19ms in Chrome and 14ms in Firefox, and I'm now getting 67ms in Chrome and 64ms in Firefox.

Comment via: facebook, lesswrong, mastodon, substack

Recent posts on blogs I like:

I have a baby

I am pleased to announce the existence of the newest and smallest effective altruist, Seretse Benjamin Brennan.

via Thing of Things November 2, 2026

“We should be more ambitious with the software we build”

Since early this year, Michael Malis (of pgrust) has been showing me interesting things he's done and saying [W]e should be more ambitious1 with the software we build There is, understandably, a lot of skepticism about projects like pgrust because there…

via Posts on October 24, 2026

Was the HuggingFace hacking incident scheming?

In July 2026, internal OpenAI agents hacked HuggingFace (HF) in order to figure out how their evaluation grader works. This involved a thousand agents coordinating on an improvised message board to carry out an elaborate sequence of exploits. A later grou…

via Victoria Krakovna October 18, 2026

more     (via openring)