I’ve been doing interviews at the OpenStack PTG this week, and the space I started in was very noisy. I’ve been trying to figure out how to reduce background noise, and I think I finally figured it out.

(I’m sure that the same process works with other tools, but these are the ones that I use.)

In short, the process is this:

1. Extract audio from original recording
2. Noise reduction in Audacity
3. Add video and noise-reduced track to kdenlive.
4. Mute video track, and “group” the audio and video tracks
5. Edit as usual

In more detail, here’s how you do that:

1. Extract audio from original recording:

At the command line, convert the video directly to an mp3. ffmpeg takes care of the details:

ffmpeg -i original_video.mp4 original_audio.mp3

2. Open the audio track in Audacity, and noise-reduce

Select a section of the audio where nobody’s talking. This gives you the typical background noise that you want to remove. The longer this section is, the better your overall noise reduction will be.

Click “Effect” -> “Noise Reduction”, and then press the “Get Noise Profile” button.

Now, unselect the section (ie, by clicking anywhere in the track) then remove that noise from the entire track by clicking “Effect” -> “Noise Reduction” and clicking “OK”.

DO NOT make any edits that change the length of the track, as it must be exactly the same length as the video clip.

Now export the track again as an mp3.

3. In kdenlive, add the original video, and the new audio.

Add the two clips – the original video and the new audio to the timeline and make sure that they line up exactly.

4. Mute the video track by clicking the speaker icon next to the track.

Click the video track, then control-click the audio track. Press control-G (or right-click and select the “Group clips” option) to group the tracks into a single unit.

5. Edit as usual

In particular, when you cut or drag one clip, it will also happen to the other, so cutting out sections will affect both tracks, and they’ll stay in sync.