jliljebl / flowblade-forum

Flowblade user forum
GNU General Public License v3.0
8 stars 3 forks source link

How to use the audio 'Normalize' filter? #85

Open CWalkden opened 3 years ago

CWalkden commented 3 years ago

Using Flowblade 2.6.1 on Ubuntu 20.04. I have two sources of sound in my videos, one is a video recorder and one is a separate sound recorder. I sometimes use one and sometimes the other, swapping back and forth for different sections of the same video. The sound levels from the devices are different, and I'd like to normalise the levels across the whole video. It looks like the 'Normalize' filter might do this? But I can't figure out how it works.

There are four sliders in this filter, but I can't find a description of what they do. Blend Mode: Adjusting this makes the audio get either softer or louder over the clip. Would be good for fading in or out. Analysis Length: I can't figure out what this does. Max Gain: Looks like it should be the maximum gain for the clip in dB, but seems to be reversed - higher numbers cause the audio to be louder. Min Gain: I can't figure out what this does.

With my fiddling I've been able to get the audio of a single clip to get either louder or softer over the course of the clip, but I can't get it to move uniformly louder or softer. Can someone explain the functionality of this filter, or point me towards where it is explained?

jliljebl commented 3 years ago

The documentation for the filter is here: https://www.mltframework.org/plugins/FilterDynamic_loudness/

I have not myself used this a lot but if there are people knowledgeable with stuff like, please feel free to offer opinion.

CWalkden commented 3 years ago

The documentation for the filter is here: https://www.mltframework.org/plugins/FilterDynamic_loudness/

Thank you, this is useful. I've done some experimentation today and found what could be a bug: Blend Mode in Flowblade goes from -50 to -10. A setting of 23 means that the volume is the same for the duration of the clip, and this is the default. Higher setting means that the volume gets louder over the clip, a lower number means that it gets quieter over the clip. target_loudness in the filter documentation goes from -50 to -10, and is a number that sets the loudness of the output of the filter (in LUFS).

So it seems that Blend Mode is probably supposed to be mapped onto target_loudness. But for some reason it is being applied on a sliding scale, and this is not useful. I think this function would be exactly what I'm after, if it worked as per the description in the filter documentation.