Autogain and severe overshoot when adding a new stream.

[ak2766]

First, I love EasyEffects. Thanks for making this fantastic and feature rich audio equalizer that makes my Linux desktop really awesome - 💖.

TL;DR

I believe your answer @ #2477 (comment) is the right path to follow to resolve this issue, but please read on as I having one question at the end.

Now the problem

PS: this was all before I had seen your previous answer regarding the compressor effect and felt that removing it all might be a disservice to others with a similar concern.

I normally have my music streaming from either Rhythmbox or Radio Garden. I normally set the volume for both quite low - the one below is from Radio Garden:

Main Problem:

If I happen to launch a YouTube video, the initial YouTube audio is too loud and distorted before autogain kicks in and lowers it. What I'm after is - low audio volume from my music stream (Radio Garden here) and loud(er) audio volume for the YouTube video (somewhat like pipe music in a restaurant):

What can I do to limit this overshoot from happening? I'm sometimes using a headset and that really hurts my puny ear drums.

OS: Ubuntu 24.04.1 (fully patched)

$ > sudo apt update 2>/dev/null | grep upgradable && apt list --upgradable && phased
6 packages can be upgraded. Run 'apt list --upgradable' to see them.
Listing... Done
grub-common/noble-updates 2.12-1ubuntu7.1 amd64 [upgradable from: 2.12-1ubuntu7]
grub-efi-amd64-bin/noble-updates 2.12-1ubuntu7.1 amd64 [upgradable from: 2.12-1ubuntu7]
grub-efi-amd64-signed/noble-updates 1.202.2+2.12-1ubuntu7.1 amd64 [upgradable from: 1.202+2.12-1ubuntu7]
grub-efi-amd64/noble-updates 2.12-1ubuntu7.1 amd64 [upgradable from: 2.12-1ubuntu7]
grub-emu/noble-updates 2.12-1ubuntu7.1 amd64 [upgradable from: 2.12-1ubuntu7]
grub2-common/noble-updates 2.12-1ubuntu7.1 amd64 [upgradable from: 2.12-1ubuntu7]
Listing...
grub-common
     2.12-1ubuntu7.1 500 (phased 39%)
grub-efi-amd64-bin
     2.12-1ubuntu7.1 500 (phased 39%)
grub-efi-amd64-signed
     1.202.2+2.12-1ubuntu7.1 500 (phased 39%)
grub-efi-amd64
     2.12-1ubuntu7.1 500 (phased 39%)
grub-emu
     2.12-1ubuntu7.1 500 (phased 39%)
grub2-common
     2.12-1ubuntu7.1 500 (phased 39%)

EasyEffect version: 7.1.6-1ubuntu0.24.04.1

Laptop Audio:

Audio:
  Device-1: Intel Alder Lake PCH-P High Definition Audio
    driver: sof-audio-pci-intel-tgl
  Device-2: NVIDIA driver: snd_hda_intel
  Device-3: Logitech C920 HD Pro Webcam driver: snd-usb-audio,uvcvideo
    type: USB
  Device-4: Lenovo ThinkPad Thunderbolt 4 Dock USB Audio
    driver: hid-generic,snd-usb-audio,usbhid type: USB
  API: ALSA v: k6.8.0-51-lowlatency status: kernel-api
  Server-1: PipeWire v: 1.0.5 status: active

Autogain setting 1 stream (Only the Radio Garden audio stream is currently playing):

Autogain setting 2 streams (with both Garden Radio and YouTube):

Potential Solution:

Your answer @ #2477 (comment) appears to be what I'm after - the compressor effect. However, when I turn it on, it ramps up the other stream too soon - meaning that small pauses in the YouTube video cause my Radio Garden audio stream to ramp back up leading to these very annoying periodic gain overshoots. In fact, I'd say they intrude more than that initial overshoot when it isn't turned on!

"Can anything be done about these?"

Compressor settings:

[wwmm]

What I'm after is - low audio volume from my music stream (Radio Garden here) and loud(er) audio volume for the YouTube video

This can be tough to achieve. The reason is EasyEffects does not process each player independently. Their audio content is mixed in the virtual device. And then sent to the filters. So the filters are suddenly going from very low volume audio to high volume.

Looking at your images I see that you have the autogain reference set to average the short-term and the long-term loudness estimation. This is not going to react in time to fast changes in audio levels. You need to use a reference that takes into account the momentary loudness M.

But this alone won't give you what you want. Assuming that the autogain reacts fast enough when using a reference that also considers the momentary loudness like Geometric Mean (MSI) what is going to happen is having both audio contents more or less on the same level without big jumps. You wouldn't have one stream on low volume and the other on a higher volume.

The compressor will also have some difficulties with this. Even if everything goes perfectly well you will have both streams more or less on the same level. I think that the only way to have one of the streams always at a lower volume would be through independent processing of each of them.

[ak2766]

The compressor will also have some difficulties with this. Even if everything goes perfectly well you will have both streams more or less on the same level. I think that the only way to have one of the streams always at a lower volume would be through independent processing of each of them.

Interestingly, w/ and w/o autogain enabled, the Radio Garden stream volume goes to near inaudible while the YT audio stream takes over focus - does that mean my experience is different from yours? But I did just now learn that the streams are mixed together. I always thought they were independent of each other - just going off of the dialog screens.

I'll keep experimenting and see if I can get what I want.

In playing around with the Attack and Release Time in the compressor, I noticed that if I wanted to boost Release time to say 1200.00 ms, I'd have to hold the up/down arrows for extended periods. Playing around further, I noticed that I can use CTRL+pageup/down and that increases by 5000 ms. However, there was nothing I found to increase by 100's. Typing in the entry doesn't take - 🤔!

[wwmm]

Interestingly, w/ and w/o autogain enabled, the Radio Garden stream volume goes to near inaudible while the YT audio stream takes over focus - does that mean my experience is different from yours?

Maybe. I use the silence threshold at its default value of -70 dB. But in your image you have it set to -50. If the estimated loudness level is below that the autogain is not going to apply gain correction because it will consider the audio to be silence. Maybe that is why Radio Garden levels remains near inaudible levels.