Podcast pony: normalisation/compression? March 5, 2012 3:46 AM Subscribe
Podcast pony: A smidge of normalisation/compression to even out the volume?
I love the podcast, but I often find myself listening with my hand on the volume control. There's a huge difference in volume between the loud moments (usually laughter or when the mods talk over each other) and the quiet moments when only one mod is talking. It makes it hard to find a volume setting where the podcast is always audible yet never too loud. A look at the waveform in Audacity shows very high peaks and a lot of variation. A bit of normalisation or dynamic range compression could even these out. Cortex, would you consider running a filter over the podcast before uploading it?
(I know DRC is a controversial thing in music, so I understand if people don't want it applied to the tracks. But for spoken word, some degree of compression really aides comprehension and listener comfort. Thoughts?)
I love the podcast, but I often find myself listening with my hand on the volume control. There's a huge difference in volume between the loud moments (usually laughter or when the mods talk over each other) and the quiet moments when only one mod is talking. It makes it hard to find a volume setting where the podcast is always audible yet never too loud. A look at the waveform in Audacity shows very high peaks and a lot of variation. A bit of normalisation or dynamic range compression could even these out. Cortex, would you consider running a filter over the podcast before uploading it?
(I know DRC is a controversial thing in music, so I understand if people don't want it applied to the tracks. But for spoken word, some degree of compression really aides comprehension and listener comfort. Thoughts?)
A lot of low-budget podcasts swear by the free application "The Levelator" for this purpose. It might be described as a combined sort-of-smart compressor/normalizer/limiter for audio post-production.
posted by RichardP at 5:07 AM on March 5, 2012 [1 favorite]
posted by RichardP at 5:07 AM on March 5, 2012 [1 favorite]
I used to do a lot of audio stuff for a community center and I was always pushing for spending funds on a compressor/normalizer/limiter and I always got pushback because compressors aren't as sexy as reverbs or speakers of eqs. "What do they do?" I say, "Oh, they are really important, they make things louder and softer" "Oh, just use the volume then, what we need a cd player/eq/electric guitar!" Arg! Not an open mic goes by when I wished I had a compressor.
Now that I think about it, this isn't just a problem in the metafilter podcast but EVERYONE on youtube.
posted by fuq at 5:15 AM on March 5, 2012 [1 favorite]
Now that I think about it, this isn't just a problem in the metafilter podcast but EVERYONE on youtube.
posted by fuq at 5:15 AM on March 5, 2012 [1 favorite]
Compressors cost nothing. You can run it through Audacity for free.
posted by empath at 5:58 AM on March 5, 2012
posted by empath at 5:58 AM on March 5, 2012
You FOOL! Compress the mods, would you? Level their brilliant peaks, and stop up their profound depths? YOU MAKE ME GAG WITH HATE-VOMIT.
Everyone knows that the dynamic range of the mods is their most wonderful and well-loved feature. Starting with a quiet, barely audible whisper of wisdom, the divine mods weave together their voices, singing the very fabric of MetaFilter into being - until they reach an overwhelming crescendo of transcendental insight, and, with a final, cosmic chord, as loud as a supernova, they unveil the funamental harmonic nature of the MeFi universe!
And yet you dare to suggest compromising that infinite and numinous song, for your own selfish reasons, into some discordant wodge of homogenised gunk? Have you not read the Silmarillion? This kind of shit does NOT go well in the long-run, Melko. Just let Mathowie Ilúvatar handle the DJing and save yourself a heap of trouble - believe me.
posted by the quidnunc kid at 6:45 AM on March 5, 2012 [4 favorites]
Everyone knows that the dynamic range of the mods is their most wonderful and well-loved feature. Starting with a quiet, barely audible whisper of wisdom, the divine mods weave together their voices, singing the very fabric of MetaFilter into being - until they reach an overwhelming crescendo of transcendental insight, and, with a final, cosmic chord, as loud as a supernova, they unveil the funamental harmonic nature of the MeFi universe!
And yet you dare to suggest compromising that infinite and numinous song, for your own selfish reasons, into some discordant wodge of homogenised gunk? Have you not read the Silmarillion? This kind of shit does NOT go well in the long-run, Melko. Just let Mathowie Ilúvatar handle the DJing and save yourself a heap of trouble - believe me.
posted by the quidnunc kid at 6:45 AM on March 5, 2012 [4 favorites]
Cortex, would you consider running a filter over the podcast before uploading it?
I actually do some per-track compression on each of mine and Matt's and Jess's tracks up front to try and reduce the dynamic range and get the three of us more in level conversational territory. I've tried to be more aggressive with it on occasion but have been unhappy with the results, but that may be as much me not doing a great job of finding a good compromise level as anything.
Matt has used Levelator at one point at least, I feel like I remember really disliking the output from that process because it was such aggressive compression that the result sounded squashed and had a lot of pumping and the music clips sounded tortured after. But again maybe that's a matter of finding the right spot on the dial.
We record our bits on different setups, and so there's a couple of challenges in the compression stuff:
- more aggressively compressing the individual vocal tracks tends to accentuate the difference in quality between e.g. a nice condenser mic and a USB headset, in a way that I worry is distracting itself
- bleedthrough gets brought up in level in the process too, which leads to obnoxious echo effects in the mix when e.g. my vocal track is 200ms off from a bleedthrough of my vocal track in Jessamyn's track
- miscellaneous we-are-noisy-bumbling-nerd sounds also get brought up, to the point where editing the podcast requires not just major section cuts but also a ton of little suppressive edits to get rid of the breathing noises, typing sounds, and random background stuff that otherwise sits fairly low in the mix.
None of this is fundamentally unfixable (we could try and be more consistent about good mic hardware and proper enclosed cans, maybe), but I can tell you for sure that that last bullet in particular is a practical nightmare that massively balloons the editing process. And none of us has superb mic discipline and I don't think that's likely to change, so part of the problem is that our vocal levels just do naturally have a lot of dynamic range and at the low end of that range it runs into the same approximate territory as unwanted noise. So gating and compressing are harder than they would be in a proper silent studio with great equipment and really consistent mic discipline.
Anyway, that's the stuff I think about with this. I do agree that it'd be nicer to have a more consistent volume on this stuff, because I totally understand that people are listening to this in less than pristine audio environments a lot of the time and their noise floor is a lot higher than what I might imagine in my utopian podcast wonderland. I'll keep playing with it to see if I can find a balance that's more aggressive on the compression side without making us just sound like NOISY PEOPLE WHO NEVER STOP SMACKING THEIR LIPS AND TYPING ON THE KEYBOARD or whatever.
posted by cortex (staff) at 7:32 AM on March 5, 2012
I actually do some per-track compression on each of mine and Matt's and Jess's tracks up front to try and reduce the dynamic range and get the three of us more in level conversational territory. I've tried to be more aggressive with it on occasion but have been unhappy with the results, but that may be as much me not doing a great job of finding a good compromise level as anything.
Matt has used Levelator at one point at least, I feel like I remember really disliking the output from that process because it was such aggressive compression that the result sounded squashed and had a lot of pumping and the music clips sounded tortured after. But again maybe that's a matter of finding the right spot on the dial.
We record our bits on different setups, and so there's a couple of challenges in the compression stuff:
- more aggressively compressing the individual vocal tracks tends to accentuate the difference in quality between e.g. a nice condenser mic and a USB headset, in a way that I worry is distracting itself
- bleedthrough gets brought up in level in the process too, which leads to obnoxious echo effects in the mix when e.g. my vocal track is 200ms off from a bleedthrough of my vocal track in Jessamyn's track
- miscellaneous we-are-noisy-bumbling-nerd sounds also get brought up, to the point where editing the podcast requires not just major section cuts but also a ton of little suppressive edits to get rid of the breathing noises, typing sounds, and random background stuff that otherwise sits fairly low in the mix.
None of this is fundamentally unfixable (we could try and be more consistent about good mic hardware and proper enclosed cans, maybe), but I can tell you for sure that that last bullet in particular is a practical nightmare that massively balloons the editing process. And none of us has superb mic discipline and I don't think that's likely to change, so part of the problem is that our vocal levels just do naturally have a lot of dynamic range and at the low end of that range it runs into the same approximate territory as unwanted noise. So gating and compressing are harder than they would be in a proper silent studio with great equipment and really consistent mic discipline.
Anyway, that's the stuff I think about with this. I do agree that it'd be nicer to have a more consistent volume on this stuff, because I totally understand that people are listening to this in less than pristine audio environments a lot of the time and their noise floor is a lot higher than what I might imagine in my utopian podcast wonderland. I'll keep playing with it to see if I can find a balance that's more aggressive on the compression side without making us just sound like NOISY PEOPLE WHO NEVER STOP SMACKING THEIR LIPS AND TYPING ON THE KEYBOARD or whatever.
posted by cortex (staff) at 7:32 AM on March 5, 2012
When I go to my secret ideal room, jessamyn the noisy bumbling nerd is my spirit animal.
posted by Meatbomb at 7:59 AM on March 5, 2012 [1 favorite]
posted by Meatbomb at 7:59 AM on March 5, 2012 [1 favorite]
When I use the normalisation feature in Audacity I always run it on the final version of the whole project and not on individual tracks.
I like this idea. Because of my satellite Internet service I rarely listen to the podcasts, but when I do I have the same issues embrangled mentions, and don't dare listen with headphones on.
posted by terrapin at 8:03 AM on March 5, 2012
I like this idea. Because of my satellite Internet service I rarely listen to the podcasts, but when I do I have the same issues embrangled mentions, and don't dare listen with headphones on.
posted by terrapin at 8:03 AM on March 5, 2012
There's potential value in using compression at various stages in the editing process. I do the per-track compression because that's the stage of editing I actually have hands on during; after I've synced up and compressed and written out the individual tracks, they go to Matt and he does the actual editing work to cut out noise and snip asides and insert music breaks and such, and he could at that point do a limiting pass on the final mix as well.
The value in compressing each individual track up front is you can get closer to equitable volume levels on each voice prior to any final mix compression or limiting. That's important because it means you are less likely to get a big gulf in simultaneous voice volume levels, which compressing the final mix won't help with at all. By compressing the different tracks, everybody is at at least a more even vocal level; you don't end up having LOUD MATT quiet jessamyn crosstalk that functionally buries jessamyn's voice because of the why psychoacoustic processing works.
But, like I said, there's probably more tweaking I can try and do on the actual compression settings used for the individual tracks in the first place.
posted by cortex (staff) at 8:12 AM on March 5, 2012
The value in compressing each individual track up front is you can get closer to equitable volume levels on each voice prior to any final mix compression or limiting. That's important because it means you are less likely to get a big gulf in simultaneous voice volume levels, which compressing the final mix won't help with at all. By compressing the different tracks, everybody is at at least a more even vocal level; you don't end up having LOUD MATT quiet jessamyn crosstalk that functionally buries jessamyn's voice because of the why psychoacoustic processing works.
But, like I said, there's probably more tweaking I can try and do on the actual compression settings used for the individual tracks in the first place.
posted by cortex (staff) at 8:12 AM on March 5, 2012
Autotune the Mods? ;)
posted by zarq at 8:17 AM on March 5, 2012 [3 favorites]
posted by zarq at 8:17 AM on March 5, 2012 [3 favorites]
It is true, I am a noisy bumbling nerd.
Also, LOOK OUT when she's hungry. The last time we had lunch, I thought I was going to be wounded by shrapnel.
posted by Devils Rancher at 8:18 AM on March 5, 2012
Also, LOOK OUT when she's hungry. The last time we had lunch, I thought I was going to be wounded by shrapnel.
posted by Devils Rancher at 8:18 AM on March 5, 2012
bleedthrough gets brought up in level in the process too, which leads to obnoxious echo effects in the mix when e.g. my vocal track is 200ms off from a bleedthrough of my vocal track in Jessamyn's track
Run it through a noise gate.
posted by empath at 9:00 AM on March 5, 2012
Run it through a noise gate.
posted by empath at 9:00 AM on March 5, 2012
You should just hire Quincy Jones to produce it and be done.
posted by shakespeherian at 9:06 AM on March 5, 2012
posted by shakespeherian at 9:06 AM on March 5, 2012
See above; that we're often pretty quiet in our actual speech means gating is an iffy proposition at best for general cleanup. Takes a lot of manual attendance to make it work in some cases, to the point where just manually editing isn't a worse proposition and doesn't have the same risk of accidentally swallowing up quiet but contributional vocalizations.
Aside from that, gating is no use during segments when someone with bleed on their track is actually speaking, so you still get bleed echo during any crosstalk segments.
posted by cortex (staff) at 9:06 AM on March 5, 2012
Aside from that, gating is no use during segments when someone with bleed on their track is actually speaking, so you still get bleed echo during any crosstalk segments.
posted by cortex (staff) at 9:06 AM on March 5, 2012
The best-sounding podcast I listen to uses Audacity and Levelator, so it ought to be doable. They regularly have three or more different people taking at various times and the end result is extremely clear and listenable, on par with a professional radio production. I'm sure Jack would be glad to offer some advice if you want to get in touch with him.
posted by exogenous at 9:11 AM on March 5, 2012
posted by exogenous at 9:11 AM on March 5, 2012
You should just hire Quincy Jones to produce it and be done.
I think we're looking for more of an alternative, independent feel. Let's bring in that guy from REM on guitar, & I'll phone up my close, personal friends Mitch Froom & Bill Bottrell.
posted by Devils Rancher at 9:15 AM on March 5, 2012
I think we're looking for more of an alternative, independent feel. Let's bring in that guy from REM on guitar, & I'll phone up my close, personal friends Mitch Froom & Bill Bottrell.
posted by Devils Rancher at 9:15 AM on March 5, 2012
I have an in with Zombie John Peel.
posted by shakespeherian at 9:18 AM on March 5, 2012 [1 favorite]
posted by shakespeherian at 9:18 AM on March 5, 2012 [1 favorite]
On a serious note, Logic has a wonderful "strip silence" feature that has a variable threshold, attack & release, & it basically creates regions as opposed to being an effect. Though I'd imagine that if there's a lot of background noise -- fans, AC running, what have you, that it would be jarring to have that cutting in and out.
Close-micing everyone will probably solve more of your problems than trying to fix it in the mix.
posted by Devils Rancher at 9:18 AM on March 5, 2012
Close-micing everyone will probably solve more of your problems than trying to fix it in the mix.
posted by Devils Rancher at 9:18 AM on March 5, 2012
I second this motion. I listen to podcasts with headphones, and had to stop listening to you guys because of the "What? Huh? AAAH MY EARS ARRRRGHHHH" factor.
posted by The corpse in the library at 9:19 AM on March 5, 2012
posted by The corpse in the library at 9:19 AM on March 5, 2012
Cortex, what do you think about hosting the raw files of an episode or two and letting the community experiment with the mix? If somebody skilled comes up with a good solution they can share it with you and everyone will benefit.
posted by michaelh at 9:46 AM on March 5, 2012 [3 favorites]
posted by michaelh at 9:46 AM on March 5, 2012 [3 favorites]
I don't have a problem with that idea, though it's worth pointing out that there's a difference between "I massaged these files with tools x, y, and z" and "I massaged these files with a workflow that will be convenient for you guys to implement", so I wouldn't want anyone spending too much effort trying to do something that works as a one-off but would be potentially impractical as a modification to how the podcast actually gets put together.
For context, our existing workflow is call recording via Skype into isolated tracks (via a little Skype addon called Call Recorder, mostly just because it consistently works), then per-track syncing and compression via Garageband on my end, then those tracks written out individually and sent over to Matt for the master edit, also as far as I know in Garageband.
posted by cortex (staff) at 9:55 AM on March 5, 2012
For context, our existing workflow is call recording via Skype into isolated tracks (via a little Skype addon called Call Recorder, mostly just because it consistently works), then per-track syncing and compression via Garageband on my end, then those tracks written out individually and sent over to Matt for the master edit, also as far as I know in Garageband.
posted by cortex (staff) at 9:55 AM on March 5, 2012
I used to use Levelator for every podcast, but for some reason I stopped about a year ago because it was another 30min of work. I can go back to using it just before upload.
posted by mathowie (staff) at 10:31 AM on March 5, 2012
posted by mathowie (staff) at 10:31 AM on March 5, 2012
I don’t know about Levelator, but there are many different types of compressors and ways to use them, so if you’re not liking something about the compression it’s probably fixable.
If you don’t want to have everything compressing all time just put a limiter, or compressor with a high ratio, a hard knee and put the threshold high enough that it only hits on the loud parts. In simple terms, ratio is how much you’re compressing, knee is how kicks in (gradually=soft, all at once=hard) and threshold is just what it sounds like, at what level does it kick in. In the this setup (high ratio, hard knee, high threshold) nothing will be compressed unless it goes above a certain level, in other words, it will just grab the parts that are killing people’s ears.
If you don’t want to mess with all that then just put a limiter (which is basically a compressor with those settings) on and set the threshold to the maximum you to blast people with.
I haven’t heard your podcast, but please, everyone, put a hi pass filter on your stuff, at least at 60hz or even 90 or so if you don’t have music. Even on NPR my speakers blast the low end bump of someone bumping the table or something. So irritating, so easily solved.
posted by bongo_x at 11:52 AM on March 5, 2012 [1 favorite]
If you don’t want to have everything compressing all time just put a limiter, or compressor with a high ratio, a hard knee and put the threshold high enough that it only hits on the loud parts. In simple terms, ratio is how much you’re compressing, knee is how kicks in (gradually=soft, all at once=hard) and threshold is just what it sounds like, at what level does it kick in. In the this setup (high ratio, hard knee, high threshold) nothing will be compressed unless it goes above a certain level, in other words, it will just grab the parts that are killing people’s ears.
If you don’t want to mess with all that then just put a limiter (which is basically a compressor with those settings) on and set the threshold to the maximum you to blast people with.
I haven’t heard your podcast, but please, everyone, put a hi pass filter on your stuff, at least at 60hz or even 90 or so if you don’t have music. Even on NPR my speakers blast the low end bump of someone bumping the table or something. So irritating, so easily solved.
posted by bongo_x at 11:52 AM on March 5, 2012 [1 favorite]
Metafilter : NOISY PEOPLE WHO NEVER STOP SMACKING THEIR LIPS AND TYPING ON THE KEYBOARD
posted by crunchland at 1:25 PM on March 5, 2012 [1 favorite]
posted by crunchland at 1:25 PM on March 5, 2012 [1 favorite]
I don't think you need compression. You just need to not set the peak volume for music far, far below the peak volume for dialog. This will help reduce the incidents of THIS IS BLASTING MY FUCKING EARS OFF OW OW OW I AM YANKING MY EARBUDS OUT that occur during the musical breaks in pretty nearly every podcast
posted by ardgedee at 2:43 PM on March 5, 2012
posted by ardgedee at 2:43 PM on March 5, 2012
Oh yeah, I meant a compressor for live sound. A little more challenging than audacity can handle. Why do I get live sound and recorded sound confused? Because whatever. Y'all should be hip about this and only perform the "podcast" live, in person, for a small audiance selected by lottery.
posted by fuq at 2:59 PM on March 5, 2012 [1 favorite]
posted by fuq at 2:59 PM on March 5, 2012 [1 favorite]
For what it's worth, I had good results compressing the most recent podcast with a hard limiter, as bongo suggested. I used the DRC filter in Audacity. The aim isn't to bring everything up the same volume, noise floor and lip smacking included. It's to keep everything within a range that's comfortable to listen to. Right now the peaks are so high that simply cutting their maximum amplitude would be enough to solve the problem. I understand your concerns about the noise floor, Cortex, but right now people are turning up their volume controls to listen to the quiet bits, so they're hearing it anyway. And to be honest, that aspect of it sounds fine, it's the peaks that make my ears hurt.
posted by embrangled at 3:15 PM on March 5, 2012
posted by embrangled at 3:15 PM on March 5, 2012
cortex, it sounds like you would benefit from gating and compression earlier in the signal chain. I don't use Skype, but if it's anything like Ventrilo or Mumble, there should be a UI for each user to set up those effects on themselves. If there isn't, I recommend using one of those free programs.
That way, you would eliminate all open mic noise, and the individual tracks would come through gated and compressed. Would probably save you some time in the final mixdown, as you would only have to selective edit and mix the volumes.
posted by lazaruslong at 3:45 PM on March 5, 2012
That way, you would eliminate all open mic noise, and the individual tracks would come through gated and compressed. Would probably save you some time in the final mixdown, as you would only have to selective edit and mix the volumes.
posted by lazaruslong at 3:45 PM on March 5, 2012
Mumble is especially fantastic for automatic level detection. The setup wizard gives you green/yellow/red metering for on-mic, ambient, and off-mic levels and really nails it pretty well. That might help with the variance in hardware setups.
posted by lazaruslong at 3:47 PM on March 5, 2012
posted by lazaruslong at 3:47 PM on March 5, 2012
Hm, I wonder if it would be worth fooling around with Ventrilo. I've had pretty good luck with it in the past, but never tried to use it to generate recordings. Skype absolutely drives me batty most of the time, although at least it's free.
posted by restless_nomad (staff) at 4:03 PM on March 5, 2012
posted by restless_nomad (staff) at 4:03 PM on March 5, 2012
If the Levelator's compression is too aggressive you might try turning down the input level (just Gain in Audacity or whatever) by 6 to 10 dBs. I was actually impressed by the Levelator, it's very simple but it does the job surprisingly well (the point is that it has no dial, to be pedantic).
(Now if someone can recommend me something like the Levelator but without the noise gating aspect, I could actually use it for documentary interview dialogue!)
posted by goodnewsfortheinsane at 4:41 PM on March 5, 2012
(Now if someone can recommend me something like the Levelator but without the noise gating aspect, I could actually use it for documentary interview dialogue!)
posted by goodnewsfortheinsane at 4:41 PM on March 5, 2012
Compressors are probably the hardest thing for people getting into audio to understand, but what goodnewsfortheinsane said, if there’s too much aggressiveness going on you’re going to get the ill effects mentioned. You shouldn’t be working it that hard, if the reduction meter is working all the time, more than a few (3-6) db, you’re hitting it too hard. Back off the input, turn up the threshold, or both. You just want to take the peaks down, not work the compressor all the time.
posted by bongo_x at 5:47 PM on March 5, 2012
posted by bongo_x at 5:47 PM on March 5, 2012
Oof, cortex, it looks like you could use some help with your workflow. My suggestions:
* Use ducking in GarageBand. That should mute most of the incidental audio when someone else is talking.
* If people are talking over each other at different volumes, turn up the quiet one. It seems like you should only have to do that once per podcast.
* Do vocals-only & music-only mixes, send the vocal mix through Levelator, then recombine them.
posted by Pronoiac at 6:10 PM on March 5, 2012
* Use ducking in GarageBand. That should mute most of the incidental audio when someone else is talking.
* If people are talking over each other at different volumes, turn up the quiet one. It seems like you should only have to do that once per podcast.
* Do vocals-only & music-only mixes, send the vocal mix through Levelator, then recombine them.
posted by Pronoiac at 6:10 PM on March 5, 2012
Maybe if the mods just shout the podcast, then you could turn the volume down. Solution!
posted by Night_owl at 6:22 AM on March 6, 2012
posted by Night_owl at 6:22 AM on March 6, 2012
Use ducking in GarageBand. That should mute most of the incidental audio when someone else is talking.
That doesn't work if he's using all vocal tracks. You use ducking to lower the volume on non-vocal tracks when someone is talking/singing. Unless you're going to go through the entire mix to figure out who the primary speaker is and make them the sidechain input (something that would be an immensely time consuming process), I don't think it would be useful. You'd do just as well adjusting everyone's volume manually.
posted by empath at 6:29 AM on March 6, 2012
That doesn't work if he's using all vocal tracks. You use ducking to lower the volume on non-vocal tracks when someone is talking/singing. Unless you're going to go through the entire mix to figure out who the primary speaker is and make them the sidechain input (something that would be an immensely time consuming process), I don't think it would be useful. You'd do just as well adjusting everyone's volume manually.
posted by empath at 6:29 AM on March 6, 2012
Re: ducking - Well, darn. I hadn't actually used it, & misunderstood it.
Hm. Is there a way to use a noise gate, & manually include the useful but quiet parts? You could find those parts with a loud monitor mix.
I'm enjoying the relative simplicity of streaming right now. I'd forgotten exactly how much fiddling I did in post-production.
posted by Pronoiac at 10:13 AM on March 6, 2012
Hm. Is there a way to use a noise gate, & manually include the useful but quiet parts? You could find those parts with a loud monitor mix.
I'm enjoying the relative simplicity of streaming right now. I'd forgotten exactly how much fiddling I did in post-production.
posted by Pronoiac at 10:13 AM on March 6, 2012
« Older Does this post hold currency? | Want to know what books fellow Mefites like? Come... Newer »
You are not logged in, either login or create an account to post comments
This is a pretty good idea.
posted by double block and bleed at 4:38 AM on March 5, 2012