Tips for Mastering for Soundcloud

Hey folks. Matthew Weiss — weiss-sound.com, theproaudiofiles.com. This tutorial is going to be about uploading your music to soundcloud, and how to preserve the optimal sound quality. If you are like me, and you put your heart and soul into getting the best sounding record that you can, when you hear it played back on SoundCloud, your reaction is probably something like this... [YELLING] Because it doesn't sound good! It's converting your lossless, wonderful record into a 128kbps mp3, which is a lot of data to be removed. The problem is you can't just chop data out. You're going to end up sacrificing quality in some way, shape, or form. Now, good codecs will do that in the best way that they can, but they're going to do it in a manner that they've deemed as “Psychoacoustic,” and they incorporate these reconstruction filters that remove data from the signal, and then take the remaining data and reconstruct it in a way that feels like you've lost the least. They do a good job, actually when you consider how much data that they remove. The more data obviously, the harder those reconstruction filters are going to work, and the more removal and the more reconstruction occurs. There are two big red danger zones when it comes to sending a signal through this kind of filter. The first one is the very top end of the frequency spectrum. Because we don't really hear too much above 16kHz, the reconstruction filters are programmed to really remove most of the data up there, which is where most of the damage is going to be done. They figure this is where you can kind of hear it the least, so this is where we're going to start getting goofy. It's kind of the same thing down at the other end of the spectrum. Everything that's underneath about 25Hz they're going to be removing a lot of information there. So that's one thing to be aware of. Then the other thing is that when we start getting toward the top of the digital ceiling — like as we start getting close to 0dBFS, that's where the other mistakes are going to occur, because the reconstruction filters have to reinterpret the waveshape when they're done, and if we have a lot of wave points that are very close to the top, instead of — let's say our sine wave, if here's our ceiling, our sine wave might have looked like this. Underneath the ceiling, when it sees the remaining points, it might go, “Oh, this is probably this sine wave that goes above the ceiling.”So it's going to interpolate the samples as meaning this signal is clipped. So, those are the two things that we need to be aware of when we're talking about very high frequency information and also very, very low frequency information, but primarily very high, and also when we're very close to the digital ceiling. Alright, so now let's take a look at some examples here. First, I'm going to start with the unencoded version of a master that I created. [mix plays] Now, I'm going to switch to a codec, which is going to be a 128kbps MP3, rendering in the middle. Here we go. [mix, 128kbps] Uh, just A/Bing those, and obviously I went through that very quickly, but the differences are that — well, first of all my mix holds up, generally speaking, which is great. Second of all — Or I should say my master holds up — and second of all, the things I feel like I'm losing are when I switch over to the codec, I feel like everything feels a little bit thinner, the snare loses some of it's punch, and there's a quality to the sound that I would describe as like a film. I would say it's like the sonic equivalent of a slightly frosted window. There's just sort of this broadband thing that's happening that's not so great, but overall not too bad. So now what I'm going to do is I'm going to play the mastered version, ran through the codec, and then I'm going to play the mastered version ran through the codec with some other adjustments, and we're going to A/B between those real quick. [mix] So I think the big thing that's most apparent is that the body of the snare comes back significantly, and also the fullness of the record returns somewhat. So what are the things that are going on here to make that difference?So first of all, just in general, when I master a record, I'm usually fairly conservative with the stuff that's happening above 16kHz. So that did not change, because I had already done some level attenuation with an EQ above — from 20kHz, which has a slope that kind of rolls things off in the very super treble range. So if you happen to be producing an EDM record in particular, where EDM you tend to have very extended top ends. Things with a lot of information above 16kHz. You might want to make an EQ adjustment when you are making a version for SoundCloud. Just a thought. I didn't have to do that. I didn't feel like it was necessary. What I did do — and this is not something I would've thought of, but I happened to stumble across an article on the iZotope website — and they said one of the things that might help is to narrow the highest frequency band. So I took their multi-band imager, and everything that's above 8.5kHz, I'm narrowing that by about 14%. I don't know the science behind that, I don't know why it seems to work, but it certainly did. So that's one adjustment that I made. Another adjustment that I made is that normally when I print, and the way I printed this final master, I have this final output level, and I set it very close to the digital ceiling. So it's at -0.03dB. It's just up against the digital ceiling with a little bit of space there, but what I did for this second print is I turned the overall output level down to -0.6 dB. So there's now half a decibel of space between where the final output of this record is and where the digital ceiling is. So I've left myself some more space for the reconstruction filters — basically to goof and not cause damage. The other adjustments that I made where I slowed down the action of the limiter. So originally, my settings looked like this. This is my initial release, and this is my slower release. This is a fairly fast acting limiter here. I didn't have the look ahead on either. So this limiter is acting very quickly. So what I did was I started the initial release faster and I slowed down the total release, and I also put the look ahead up to about 0.25 milliseconds, and ultimately without getting too much into it, I'm slowing the limiter down so it's not moving as quickly, and that reduces the amount of what's called intermodulation, which is something that does not play nicely with reconstruction filters. That's how I got this A/B. [mix plays] So then I wanted to take it a step further. I said, “Can I get an even better quality sound if I'm going to go to SoundCloud?”I came up with this. [mix plays, mastered for SoundCloud] And the way I did that was actually sort of counterintuitively by taking the amount of level that was feeding the limiter and turning it down by half a dB. So instead of doing about 5 dB of level attenuation, I'm now only doing about 4.5 dB of level attenuation. In other words, I'm making the limiter work less hard, and so yes, I get less playback level, but to be quite honest, it sounds significantly better when I go between the two. [mix playback] And remember, the end listener has control over the level of the playback, but the end listener does not have control over the quality of the playback. That's very very important to remember. So if you're doing a style of music that generally speaking is very heavily compressed in terms of it's dynamics: Metal, EDM, Hip-Hop, Pop, pretty much everything these days, when you make a version for SoundCloud, be more conservative on the playback level. You will get a better quality of playback, and at the end of the day, I think that the better quality is ultimately what's going to win over your listener. Alright guys, until next time.