Audacity Forum

Hi all,

I'm trying to get this recording of an orchestra to sound more professional (i.e. like a soundtrack):

MP3

The obvious problems are:

1. It's too soft overall (-> needs to be amplified)
2. Dynamic contrast is too great (-> compressed)
3. It's too "dry" (-> reverb)
4. Sound is a little muddy (-> EQ?)


Can anyone fiddle with this track and get a better result?














Below, I tried to get it closer to my ideal sound, but it's not much closer:

Revised MP3


I don't have any Audacity know-how other than what I learned fiddling around with the effects, here's what I did.

STEP 1. EQUALIZATION
I plotted three points:
100 Hz: -3 dB
1200 Hz: +3 dB
10k Hz: 0 dB
with smooth slopes between them.
This created a bit brighter sound and differentiated the instruments.

STEP 2. REVERB
added Gverb as follows:
Roomsize: 50
Reverb time: 1.5s
Damping: 0.5
Bandwidth: 0.75
Dry signal: -1.5 dB
Early reflection: -10 dB
Tail level: -20 dB

STEP 3. DYNAMIC COMPRESSION
I applied the default settings (-32dB thresh, 2:1, 0.2 seconds attack).



As you can hear the results are not great. There seems to be a "metallic" edge on the sound and it feels like the instruments are too "close" to the ear (not sure of the technical terms). I think I overdid the reverb a bit too.

I would really appreciate advice from the pros here, on how to do a better job.

Thanks in advance for all tips.

A lot of this is about personal taste, so my preferred settings may be different from yours, but there's a few things that I would do slightly differently.

I would certainly use Audacity 1.3 rather than 1.2 as both the effects and edditing features have been considerably enhanced.
Hopefully you have an original WAV format recording to work from - MP3 is convenient for smaller files for electronic transfer, but lowers the sound quality.
Work with a 32bit track - as well as improving the processing quality, in recent versions of Audacity 1.3 it does not matter if you go slightly over 0dB during your processing.

The main thing is that I would apply dynamic compression before adding reverb. If you apply reverb first, then the dynamic compression will make the amount of reverb go up and down according to the level of the audio, so quiet parts will end up with a lot more reverb than loud parts.

First I would apply the Amplify effect to bring the peak level up to 0dB.

I would use Chris's dynamic compressor ( http://pdf23ds.net/software/dynamic-compressor/ ) rather than the default Audacity compressor - try this with the default Compress Ratio of 0.5 - higher number for more compression, lower for less. Personally I would not go above 0.6 on this type of music - probably less.

I would Eq it a bit differently. For my taste the bass has been rolled off too much. I would keep the level up at 0 at 100Hz, then dip it down between 200 and 400 Hz, then slowly bring it back up to zero at 1kHz. I would probably also add a bit of boost between 50 - 100Hz and 1kHz to 10 kHz (so there is a subtle presence peak around 2kHz). Make the settings as a reasonably smooth curve to avoid sudden transitions - but no need to go over the top with adding control points.

I tend to prefer Freeverb to G-Verb, but in either case I would agree on keeping the length reasonably short (G-verb has a ridiculously long default value). I would probably go for a slightly longer reverb than you have set - around 2 seconds - and a slightly larger room size (about 70). Rather than adding it directly to the track I would make a duplicate track and apply reverb with 100% wet (no direct signal) to the track. This has several advantages;
* You can then adjust the wet/dry mix using the Mixer Board by adjusting the relative levels of your wet and dry tracks.
* You can Eq the wet and dry tracks differently. Note there is a subtle difference between Eq'ing the reverb track before or after applying the reverb. For the reverb track I would Eq the 100Hz to 1kHz dip, but not boost the low bass or high treble, and I would apply this Eq before applying reverb.
* You can insert a small period of silence at the beginning of the reverb track so that the reverb is slightly delayed - probably no more than about 40 milliseconds. This will allow you to use more reverb without making the sound too mushy. It is a rough approximation to the common recording technique of mixing close microphones (dry) with "ambient" microphones further back in the room.
* You can Eq the reverb (after applying the reverb) - Soundtracks often have a brighter reverb than "natural", but you can create any "room colouration" that you like.

Finally I would use the Amplify effect to set the peak level just below 0dB (about -0.1).

So the order of processing:
1) Convert to 32 bit if necessary
2) Amplify to 0dB
3) Compress track
4) Duplicate track
5) Eq each track - one for the dry part of the mix, one for the wet part.
6) Apply reverb
7) Mix
8) Amplify to -0.1dB
9) Export as WAV
10) (optional) Export as MP3

Here's a link to an example - it is a bit more compressed than I would normally go for (Chris's Dynamic Compressor set to 0.6), but I'm not sure exactly what you are after.
http://www.sendspace.com/file/o7jlkt (23MB Flac format)

Steve: thanks so much for your tips (especially stuff I never thought of like putting a tiny delay on the reverb track).

This is just a rehearsal recording, I'll get a concert recording tomorrow and apply your advice. I'll also get 1.3.x and mess around with it today

EDIT: the original file is WMA (off a cd) - I used WinFF to convert to WAV (and to convert the finished Audacity file back to MP3). I see in the beta export-to-MP3 is finally an option

OK, after playing around with 1.3 and the plugins you recommended I was able to get a great sound.

Final MP3

BTW your advice to apply a second post-reverb EQ was a real insight - I just rolled off the 600Hz range again and it made the final mix very colorful.

The only remaining slight issue is one spot in the track where there is unwanted "resonance," almost a feedback-like sound. Here's a clip of that:

Feedback clip

Not sure if this is fixable.

AudioHelp wrote:EDIT: the original file is WMA (off a cd)

I'm guessing that you ripped the CD using Windows Media Player.

Ripping directly to WAV avoids the losses caused by WMA encoding.

It is possible to configure WMP to rip directly in WAV format, but a better solution is to use a better CD ripper such as C-Dex or Exact Audio Copy. Both are free and eaasily found on Google. Exact Audio Copy (EAC) can provide "bit perfect" copies when ripping a good quality CD but becomes extremely slow with scratched CDs. C-Dex produces very high quality copies (but not necessarily bit for bit identical) and is faster than EAC. C-Dex handles scratched CDs much better than EAC. It is worth having both programs installed.

AudioHelp wrote:I see in the beta export-to-MP3 is finally an option

Export to MP3 has been available in all versions of Audacity for quite a few years, but requires that Lame is installed. The menu option for Exporting to MP3 is different in Audacity 1.3.x to 1.2. Audacity 1.3.x supports using VBR MP3s (Variable Bit Rate). VBR will generally produce higher quality sound for comparable bit rates to CBR (Constant Bit Rate). Most modern MP3 players will support VBR MP3s, but some may report the length of the track incorrectly (not usually a problem).

AudioHelp wrote:OK, after playing around with 1.3 and the plugins you recommended I was able to get a great sound.

Very nice

AudioHelp wrote:The only remaining slight issue is one spot in the track where there is unwanted "resonance," almost a feedback-like sound. Here's a clip of that:

Oh dear, that's a bit weird.
I think that you may need to try and isolate at what stage the problem occurs (which unfortunately may involve going through the whole procedure again).

I suspect that the problem is either caused by the reverb, or when you mix the original with the delayed reverb.
I think what is happening is that somewhere in the overall reverb effect, there is a delayed signal that is very close to one of the dominant frequencies in that section of the music and that is creating this "ghost" note.

Listen very carefully to both the original and the reverb tracks to see if the problem exists in either track on their own. If the problem is on the reverb track you will need to try tweaking the reverb settings. If the noise only occurs when you play both tracks together you may need to tweak the amount of "delay" that you have added to the reverb track.

As a matter of interest, which reverb effect are you using?

You guessed correctly re: ripping, I'll download C-Dex now (I'm recording the dress rehearsal in a few hours ).

For that track, I used Freeverb with these settings:

Roomsize: 0.700
Damping: 30-50% don't remember exactly
Predelay: 0
High/lowpass: 0
Wet: 0
Dry: -endless

I'll listen to that spot carefully at each stage of the process, for the final recording.

re: MP3, I downloaded the lame.dll. The MP3 is just for sharing online, I'll also export a WAV for CD's.

AudioHelp wrote:For that track, I used Freeverb with these settings:

Roomsize: 0.700
Damping: 30-50% don't remember exactly
Predelay: 0
High/lowpass: 0
Wet: 0
Dry: -endless

That's interesting - the Linux version of Freeverb does not have the Predelay option. I assume that this is the same as we were doing by adding a bit of silence to the beginning of the reverb track.

Good luck with your recording.