Hi!
I intend to do some screencasting t teach my pupils to use inkscape, i've already started but i'm particularly unhappy about the audio quality i get (apart from the silly sound of my voice). I've got an integrated sound card and a 10€ microphone, so i suppose i'll have to upgrade that equipment, but i don't know how. Any suggestions?
Thanks

Ah!
I'm using recormydesktop under ubuntu 7.04

A cheap sound card isn't an issue and a cheap mic is usually fine, too.

There are a few potential issues tho.

Some very cheap mics don't have foam around em. This foam acts as wind shield and is very important. Eg if you gently blow over a mic you'll get a totally maxed out (even clipped!) waveform. Now the problem are hard consonants such as 'k' or 'p'. If those are spoken a rather hefty stream of air hits the mic. Without any kind of wind shielding you get irreparable artifacts in your waveform and it will sound truly nasty. Some headsets dodge that issue by making the mic point away from you (so the air doesn't hit the mic directly). And really expensive mics have a separate big windshield, but you won't need one of those... unless you want to start a career as singer.

I recommend to do a few test runs in Audacity. Try different mic positions and different sensitivity settings. Ensure that the waveform isn't clipped anywhere (that is running flat for a bit at the very top or very bottom).

The next problem is static noise. Chances are that the environment isn't perfect for audio recording, but Audacity can help here again. Effect->Noise Removal works pretty well. Use that and all the static noise will be gone.

After removing the static noise use Effect->Normalize to normalize the volume levels. This will basically make it as loud as possible (without any clipping). Most screencasts are too quiet and that's often a bit annoying, because ramping up the volume isn't always possible. Turing the volume down, however, always is possible.

Eg I can ramp up the volume by about 20%. If I want to move to the (usually) waaay to loud level I have to walk over to the amplifier... and well, that's pretty annoying. In the past there were also a few on-board sound cards with very low output levels. If you use one of those with headphones in a noisy environment you won't be able to understand a thing, if the recording isn't at the usual volume levels.

In a nutshell:
1. ensure that no clipping occurs
2. record
3. remove noise
4. normalize

Thanks aho!
I'm testing my homemade spounge windshield, and i've already installed audacity to proceed as you tell me, but i have a problem, i cant't import properly the video files that come from recordmydesktop (ogg), they appear shortened (4 mins. video appears as a 5 secs audio file) and sound like 1000 scrathed records at a time, could it be a misconfiguration of my computer or is it that audacity can't import ogg video files?
Cheers.

Ogg is a container format. Audacity can only deal with ogg files if they contain Vorbis stuff.

However, Ogg Vorbis is a lossy format and you shouldn't use it for intermediate steps. The reason for that is the reduced quality of the final product. Or more precisely: a degenerated quality/size ratio. You see... lossy transformations introduce artifacts which are very hard to encode, which means that the quality degenerates even if you use higher and higher bitrates.

The solution is quite simple (and the same for images, audio or video): use lossless formats till the very end. Once you're ready to put it online use lossy formats/codecs to reduce the file size. This gives you the best possible quality/size ratio.

Video
Use a lossless RGB codec. It should be an RGB one to avoid color space transformations, which may cause color drifting with each step (eg white can get darker and darker with each encode/decode step). Additionally, the codec should be fast - file size isn't a concern here. The one I'm using is CamStudio in its LZO mode. Be also sure to use dimensions which are a multiple of 16. Some lossy codecs work best with 16x16 cells and some encoders only work with 16x16 cells.

At the end...
proprietary: H.264
open: Theora
(use 2pass encoding for better results)

Audio
44.1 khz 16 bit mono WAV

At the end...
proprietary: HE-AAC
open: Vorbis or Speex (if it's speech only)

One ton of thanks to you aho for all the explanations and the work you're taking to help me. I will take your advise and try camstudio, and let you know.

When I installed my tablet on Vista I noticed a really sweet feature that when performing a mouse click using the tablet pen, it was visually indicated on the screen by a ripple effect around the cursor - a nice idea for tablets since the nibs don't have a tactile click.

Anyway, I thought I'd look around to see if I could get that effect happening for normal mouse use and found SpotOnTheMouse. It's not sexy enough for everyday use, but it should prove to be very useful for the screencasters among us, since it provides visualisation of all mouse and keyboard input. Sorry, Windows only.