Compressing Sound for Videos
I got this very useful advice from Fredreyk today, he suggests that if I do videos that I should use a process that improves the sound quality so people can hear what I’m saying. It’s the same thing that adverts use to make themselves sound louder compared to the show your watching: compression.
To do this you need to take your recorded video (in my case an ogv) and using Avidemux-gtk you strip out the sound and save it in it’s own file by going in the Audio-Menu to save the audio-stream.
Now open the audio file in Audacity mark everythink(ctrl-a) and use compression in the effect-menu i took -44db in the first slider,
but it depends on your material. Just play around, so that you stop sounding like darth vader. Save the results.
If the file sounds fine then put it in avidemux together with the video-stream (open the first point (main…) in the audio-menu and use “external”). Double checking that the sound still syncs with the video.
Tags: sound editing, Ubuntu, video editing
Not that it’s worth much, but here’s what I do. Because I can’t seem to get recordmydesktop working properly with my microphone, all I do is record video using recordmydesktop and audio using “Sound Recorder” – I think it’s in the repositories. A lot of people have difficulties recording video and audio at the same time using recordmydesktop. This solves that problem.
After I’ve done with those two, I save the sound from sound recorder as an mp3 and obviously recordmydesktop saves an ogv after you finish recording automatically.
Then, you can import the mp3 audio into Audacity, apply your compression and then I personally use Pitivi to combine the two. Why Pitivi?
1) I like it more than Avidemux – it’s easier to use
2) It has more rendering options – well, i’m not sure about “more” but it certainly seems easy to choose lots of different outputs.
and 3) It’s available in Lucid by default, so I don’t have to install another application.
I then make sure everything’s lined up, cut out unnecessary bits (like when something is taking ages to download) and I cut the first 2 seconds and the last 2 seconds. All this is very easy in Pitivi.
Then I render it to 720p and upload to YouTube.
Benjamin, you’ll probably be interested in this article:
http://doctormo.wordpress.com/2009/11/18/recordmydesktop-cheesed-sound/
It explains my own trials with getting gtk-recordmydesktop to work.
You can further improve the sound quality of your video by reducing low-frequency noise that’s caused by fast-moving air hitting your microphone’s membrane. There are several occurrences of such noise in your Ground Control demonstration video (it’s apparent each time you breathe or speak directly into the microphone). It’s especially bothersome if you listen to the video on a decent set of speakers or if you have a subwoofer.
The problem can be reduced a lot if you apply an equalizer to your sound file to take out all low frequencies below about 50 Hz (these do not occur naturally in human voices anyway, unless you have an exceptionally deep voice like Barry White, I guess). Make sure to experiment with the cut-off frequency.
For best results, apply this equalizer before compressing the sound.
This technique is basically a poor-man’s emulation of a pop filter (see Wikipedia for description).
From the low threshold you mention and the sound of your Ground Control video I don’t think you’re doing a great deal of compressing but just raising the volume. Do you leave the make up gain box checked? And is the ratio fader set at the default 2:1 as that is a very gentle setting? 4:1 would be a more usual starting point.
Compressing increases the level of background noise so it can make things harder to understand so depending on the environment you record in it might be worth trying other things first.
Although I found your video perfectly intelligible there are a couple of things that I think may improve it without reaching for excessive dynamics control.
Now whoever thought up the phrase PPA has never had to deal with recording explosive consonants as when you say that there is a noticeable blast of air into the microphone. The same ‘popping’ sound can also be heard elsewhere as can the occasional movement of the microphone on your neck.
As they are far louder than you talking if you’re setting your own recording levels that would either force you to record too quietly or put up with distortion, whilst on cameras and software with automatic level control you get unnatural pumping. Also the compressor acts most on the loudest sounds so settings that deal with pops won’t do much for your voice.
So if you can angle the mic out of the air flow and secure it so it doesn’t move you can increase your recording gain if you have control over it. (-6 is usually a safe compromise for voice on a digital system at 16bits).
It’s not necessarily applicable here but speaking a bit louder would improve your signal to noise ratio and hanging a duvet behind you reduces the effect the sound of the room has on your recording.
As the voice doesn’t have a great deal of bass but pops and rumbles do it can be useful to roll off (turn down) low frequencies. Using the Audacity equaliser in graphic eq. mode you could try just pulling down faders from the left until you can hear it make a difference to your voice.
The same applies at the other end with the high frequencies if you get a horrible whine try pulling out a few faders from the right. If you end up sounding like a telephone bear in mind that it was for telecommunications that eq, compressors and more exotically vocoders, were invented.
Using a plugin that you can change in real time would help here as you can hear straight away what your changes achieve.
These days Adverts and music tend to use an extreme form of compression, limiting, to create a ‘brick wall’ threshold above which the sound can’t pass. So when the sound reaches the level at which the compressor starts working (threshold), instead of the sound only turning up by the amount set by the ratio, eg. 2:1, for every two dB increase in gain the sound will get 1 dB louder, it stops it getting any louder at all. You then ‘make up’ the difference by increasing the volume afterwards.
With modern technology it’s possible to do this fairly transparently but the temptation is to raise levels to the point where you start cutting off parts of the recording which creates distortion and leads to the harsh sound in a lot of music today.
But it might be worth trying some limiting to raise the overall level without changing the character of your voice and increasing the relative amount of noise.
Avoid setting the final level of your recording beyond about -1dB as inaccuracies in metering combined with the poor performance of mass produced digital to analogue converters can result in clipping well before 0dB.
I hope that’s more helpful than confusing.
http://www.soundonsound.com has a lot of very useful articles on recording pitched at a range of levels of experience.
For those wanting to record better screencasts on Linux with high quality audio and video, see here:
http://verb3k.wordpress.com/2010/01/26/how-to-do-proper-screencasts-on-linux/
Oh ho ho ho, an article for everything xD
Thanks!
Sadly avidemux-gtk refuses to open ogv or ogg files
Not exactly the subject matter but:
listening to your last video, once or twice you moved and spoke closer from the microphone. The sound was better then. Try to wear the headset properly instead of having it around your neck?
Also I wanted to increase the volume from your video but couldn’t. Try using a web video player that allows to increase/decrease the sound volume?
I was just asking hubby yesterday, after the ‘big install’ and my entrance into the world of Linux “Have they any good video/sound editing software?”
I used to play around a bit with Adobe Premiere, but would love an easy to use editing software in Linux/Ubuntu. I have on the back burner, as part of my continuing Vintage Project, to write, direct and produce small vintage videos based on 1950’s style tv/movies to put on a future channel on Youtube. As hubby does not care as much about video, perhaps you could recommend a good program to use. Thanks and nice to ‘meet’ you on my blog.