I have a recording in which there’s a large difference in volume between the two participants. I attached a short sample (the whole video is 50 minutes long and the two presenters switch a few times but the one with too low volume speaks during the majority of the clip). To avoid recording again, it would be great to fix this difference in Kdenlive. I tried using the Normalize filter with default settings but it didn’t really help here. What else can I do? Is it even doable?
Normalizing is not the right effect to match such different loud passages. Normalize raises the volume to a desired level, usually amplified, but of course quite evenly, for all sounds, loud and quiet. This is of no use to you in your case.
With a compressor you will have more success, because it reduces (compresses) the loud passages and then raises everything, including the quiet ones. This creates a smoother sound and greatly improves intelligibility. I always use a compressor for speech, even if it is an excellent recording.
The best compressor in Kdenlive is the SC4 (or SC1, SC2, SC3 or SC4 mono). Not sure if it is already present in the default installation. It is a plugin from Steve Harris’ SC4 – ladspa.1882. You have to install: Steve Harris’ LADSPA SWH-Plugins
With this setting it already sounds much better.
But the best method would be to apply the Volume (keyframable) effect, and then manually boost where the sound is too quiet. But this is a lot of work - you have to know if it is worth it.
Thank you for the reply. Meanwhile, I tried with manual edits using Volume (keyframable) and the effect is really good:
Regarding the SC4 compressor plug-in, is it possible to add it to Windows version of Kdenlive without having to compile from the source code? I’ve only found some information regardin the Appimage installation for Linux. Are there any guides for Windows?
If you use Volume (keyframable) - that’s fine. After that, you can still use a compressor. Unfortunately, I don’t know anything about Windows anymore. But it is certain that with the AppImage in Linux, for example, this SC4 is unfortunately also not included.
There is a compressor for you after all:
Compressor/Expander – Compress or expand the audio’s dynamic range. (avfilter.compand)
Even in the standard setting, it delivers breakable results.
Try it out. Then take the setting that suits you.
But before, as discussed volume (keyframable). It’s a lot of work, but you learn as you go.