Removing Noise with AI - RM Noise

After some confusion on my part, I concluded that the “bypass” option does not work the way I expected it to. Perhaps this post will save others from the same confusion.

When listening to audio from the radio, [bypass = true] means you are hearing the unprocessed audio FROM THE RADIO. This is the same as moving the filter slider to zero [EDIT: Not exactly the same.. see the next post from Randy]. However, when playing back a wav file, [bypass = true] means you are no longer listening to the wav file but instead you hear the radio (assuming it is on and connected).

The upshot for my demonstration purposes is that, with radio input, you can compare the filtered vs. unfiltered audio with either the bypass option OR by reducing the slider to zero. However, to make the same comparison with a recording you can only use the slider.

The behavior that I expected was that Bypass would bypass filtering but not change the input source selection. I guess the documentation is literally correct but perhaps unclear:

Monitor this device (server bypass)
For troubleshooting, this option bypasses the server and sends the radio’s output to the speakers or headphones. This box should be unchecked during normal operation.

Please correct me or clarify as appropriate.

Your observations are accurate. BYPASS always sends audio immediately from the INPUT DEVICE to the OUTPUT DEVICE.

For demonstration purposes using a recording, I suggest using the HOME (100% filtering) and END keys (0%) to toggle the filtering completely on/off.

With audio from the radio, there is a subtle difference between BYPASS and the FILTERED/ORIGINAL slider at 0%. BYPASS doesn't incur any network or buffering delays.


(The documentation you quoted is correct but somewhat dated. I intend to finish the multi-server client and then fix the documentation to match.)

RE:

With audio from the radio, there is a subtle difference between BYPASS and the FILTERED/ORIGINAL slider at 0%. BYPASS doesn't incur any network or buffering delays.

Ahh,, that makes perfect sense.

So I am grappling with being able to record the filtered audio from the non filtered audio on the same PC. Is there a way to record the output of what I am hearing on the PC speakers using Audacity or similar? I have tried changing the input settings on multiple recording software's but the end result is on the recording I still just hear the AF output from the rig and not what is happening after the RM filtering. Is there a way to record this for demonstrations?

Thanks for any help.

N2KI wrote: ↑Tue Mar 12, 2024 2:41 pm Is there a way to record the output of what I am hearing on the PC speakers using Audacity or similar?

I can think of 3 possibilities:

1) Audacity (this example is from version 3.4.2):
a) change "host" (windows sound api) to: Windows WASAPI
b) change "recording device" to your speakers (this isn't available on some soundcards)

2) I have no experience with it, but you might be able to do this with "virtual audio cables"

3) RM Noise recordings:
a) make a recording in RM Noise (any recording shorter than 30 minutes will be rejected from training the AI)
b) from https://ournetplace.com/rm-noise/ navigate to ACCOUNT -> RECORDINGS
c) wait up to 30 minutes for the filtering, refresh page
d) go to REJECTED RECORDINGS and select your recording
e) choose FILTERED RECORDING
f) under the spectrogram, choose the 3 dots to the right of the volume icon
g) choose DOWNLOAD to download the filtered recording using your browser

[edit]
I thought of another way.

4) Extract the audio from a windows screen capture:
a) start screen capture (windows-key + G, record)
b) use a tool like ffmpeg to demux the audio from the captured video
[/edit]