Skip to content

Releases: airwindows/Meter

v0.5.5

Choose a tag to compare

@github-actions github-actions released this 10 Oct 13:16

Refinements for this release don't involve changes to the metering or the black line, instead I'm LABELING them in a way that helps understanding (also added a label to the peaks section of the meter)

The labels for Detail (with the black line) now read Exciting, Present and Absent, with an extra label across the top that reads "top edge means enough intensity to be alarming and scary". That's really what's going on with the line: it maps very well to 'lead vocal characteristics' but it's really tracking the overall energies of the whole track, meaning it's just as valid for electronic genres and things with no vocal at all.

If the black line's lower than Absent, expect the track to go nowhere. Either the dynamics are so great that you're doing a big lull in the energy, or the sounds are so soft that there's nothing commanding about them (this includes vocals). This can be remedied simply by turning stuff up: nothing modern will sit around in this zone unless it's doing very quiet things for contrast.

In the region between Absent and Present, the track will be mellow, soothing. An early idea for this label was Comfort. The idea here is if you've not had the line rise to Present, it sure as hell isn't a hit record and might not merit active listening. Pretty much anything that's trying to be music to listen to, will reach Present at key points, and most things trying to be interesting start out at least at Present.

In the region between Present and Exciting, this is where popular music lives. Music does not operate by attacking the listener alone: there's a whole range of musics and energies to MATCH the vibe of the listener, to amuse or entrance or hype up, and the black line in Meter does a pretty good job of tracking the energy vibe whether it's a human voice, a guitar tone, or a synthesized noise. We react to all that stuff as if it's an emotive display, so we're alerted by dramatic yawps and soothed by comforting noises. If the line's reached Present, there's something going on, and the degree of engagement goes up as the line rises. The music becomes more demanding, harder to ignore. If you're shooting for soothing, that's a mistake, but if you're trying to make hit records there's the next region…

In the region between Exciting and the top of the detail meter, this is where the hit records live. Even mellower stuff like big poppy Journey hits pointedly go into this region: you'll plainly see Steve Perry money notes nail the middle of this region, even when the rest of the song isn't as demanding, such as 'Don't Stop Believin'. You'll also see Neal Schon's arresting solo hook do the exact same thing. For mainstream pop reach, this area is key but you won't see the black line go right to the top of the meter, instead it's all about how much time it spends hanging around in this zone, and how broad the line is (the broadness indicating sonority and how intensely the overall track is putting forth that particular energy). Remember, it's tracking slope, so it's got nothing to do with frequency, that can happen at any slope. It's more to do with how loud frequencies are relative to each other, it's a measure of how shrill everything gets.

And at the top of the meter, we're in 'Helter Skelter' land. That's not a joke, literally if you play the original White Album record and Helter Skelter, it hangs out exactly at the top of this meter for the entire time, without much in the way of loud cymbals to complicate the issue. It's just plain screaming the entire time, in a manner fit to scare anybody, and that's why the label up there reads 'enough intensity to be scary' and is centered so you risk covering it up with the line. It's too much, it's not pleasure listening, it's not even power and potency anymore, just screeching. If you try to overcook 'hit record hype!' you end up here and trade off success for desperation. This is fine and valid if that's your intent, to be 'Helter Skelter' in spirit, but if the idea is to maximize your music what you must do is TARGET the area that fits the song you're doing, and the way you're presenting it.

If you push even farther than that, the line collapses and you lose it, so even if you want Helter Skelter spirit it's still possible to dial in the mayhem and loudenation to maximize it. This should also serve the needs of overloudenators: if you can't maintain the black line, you are losing energy to distortion and making your stuff weak and bloated, so choose wisely.

The Peak meter now has a label at the top reading, "top edge means the output is clipping or beyond clipping". Clips are still shown by red markings at the bottom of the meter. My idea there is still that on the peak meter, you'll get the best sound even when going for ultra-loud by making a cloud of peak dots nearer and nearer to the top without pinning it to exactly 0dB, so what I'd suggest is to keep that narrow band where the text is, readable even if you're kinda overlapping it. Anytime your peaks are so dense up there that the text can't be read, you should back it off just to get a better texture on whatever you're trying to do, because anything covering up that text is always going to be so distorted that you'll be losing more than you gain.

v0.5.4

Choose a tag to compare

@github-actions github-actions released this 08 Oct 23:54

This version of Meter introduces a whole new measurement, except it's part of Detail (which is based on slew rates), but it's an entirely new thing for Meter that involves tracking EVERY slew rather than just keeping track of the largest one in a section!

I've got an old plugin called Beam. It's another dither plugin, with a twist: it tries to dither everything so the audio's either going up by a given slope, or down by a given slope. You're meant to dial in the degree of angle you want all the waveforms to have, and it kinda worked but was very hard to interpret what it was even doing, even at very low bit depths.

Meter now takes ideas from that, and measures two things: the average angle of the audio, and the amount of departure from that angle. It does this on the fly and tracks it, rather than only using one setting, and it draws it on the Detail meter as a black line. Sometimes, a big heavy obvious black line. And then it works that into the hit record number score setting.

If you look at the detail meter, you see it has lines on it like the other two, though they're not labeled as there's no specific meaning for where slews go. Well, now there's a purpose! The top area (what would be 0 to -6 dB on the other meters) represents a sort of intensity zone for the black line. That line will go higher on the screen as the 'average angle' gets steeper in the audio, and it will be bolder and thicker when there's not a lot of stuff distracting from and being a departure from that angle.

This becomes a way to voice things like lead vocals, and to measure them for hit record qualities. If a vocal hits the 'money note' you'll likely see it show up as a heavy black line in that top region. Down near what would be the -6dB line is the border between really mellow and low energy, and a more passionate vocal presence. Halfway to the top is where you find the memorable stuff, key hooks, the chorus. Up at the top is where you find Helter Skelter :) quite literally, in fact. But also, certain super-intense 80s tracks and the like. Go harder and the black line will be squished against the top, harder still (or into loudenation and excessive brightness) and the line will tend to disappear: it needs waveform angles to exist, and clipping tends to obliterate them.

The black line (and how bold it is: the weight of the line also matters) shows how well your hit record is communicating its lead vocal or key hooks. You might find interesting correlations between where it goes, where the red dots in Detail land, how high it goes for key moments, remembering that the position of the black line isn't necessarily tied to the slew dots in Detail. You can have a relatively dark track and a super-intense black line, or a bright and airy track and nothing much going on with the black line. It depends on your lead vocal, how intense it is, and how intense it ought to be. :)

v0.5.3

Choose a tag to compare

@github-actions github-actions released this 08 Oct 21:18

An interim post done just after my livestream…

v0.5.2

Choose a tag to compare

@github-actions github-actions released this 03 Oct 10:55

The only change here is, instead of possibly assigning a number from 1 to 99, Meter now does a Double Top 40, as in 1 to 80. Not 99, 80. That's the only change. I was measuring some Deadmau5 tracks and seeing which ones rated real high, before moving on to the massive vinyl record collection, and I decided it needed to start at 80.

Why? Because that's roughly where 'quality' (as in, actual chart success) starts to correspond more closely to the letter-based scores. There's a lot of tracks that can put up a 99 or 98, and it clutters things. At around 80 it thins WAY out, but going purely to top 40 would be too restrictive, so it's double top 40. Most of the time there won't be a number unless you're specifically going for one, but from working with it on the giant vinyl record collection with top 10, top 40 and top 100 hits flagged, there are plenty of tracks that put up a Top Double 40 number (not sequential: that's the whole problem with having a dozen '99' rated tracks) and the recent adjustments have leaned it in the direction of widely accepted 'great flawless production' tracks.

There's still 0.5.1 if you'd like numbers from 1 to 99 with no other alterations :)

v0.5.1

Choose a tag to compare

@github-actions github-actions released this 28 Sep 12:29

What I've changed in 0.5.1 is, Meter's more accepting of darker mixes. The way it counts 'detail' and the letter it assigns to detail is one letter higher than it previously used, meaning it will call things 'balanced' if they're a bit darker than before. This was done because it brought a number of titanic hit records (notably Mutt Lange's Back in Black and Pyromania) more into registering as balanced. Also works for Vanda and Young AC/DC.

To deal with the 'twitchiness' of the letter grades, those letter high scores now apply an IIR filter meaning a 'hot' response registers, but has to repeat a few times to really push the score up. I've left it reaching much the same readings as before, it's just going to be less twitchy (still not 100% repeatable though!). Along the same lines, I've totally overhauled the 'number bonus extra score' it can produce when rating things very very high: rather than being integer scores to go with the letters, it's now using double precision to bump that number up by a more smoothly varying amount that's associated with the maximum distance between the 'letter scores', and it's using the floating point versions of those scores, not just the letter grade shortcuts. So for the same letters, that number might be increasing more quickly if the underlying balance is actually more balanced. I'm hoping this will produce more reliable bonus number scores, and looking forward to seeing what Meter makes of a large and varied collection of songs.

So yeah, I'm doing all my measuring over. It will take a while, I'm up to more than 1600 songs, still stuck doing them one at a time at normal speed, but that has never stopped me before. And I think all this makes it be consistently a better and better guideline for dialing in mixes, and that's what most people will be doing if they can interpret what Meter tells them :)

v0.5.0

Choose a tag to compare

@github-actions github-actions released this 26 Sep 15:50

This changes the calculation of the 'letters' associated with each category, so that rather than bumping them up by one each time the previous one is exceeded, it's keeping a floating-point version and each time the new value exceeds the previous, it goes a fifth of the way towards the new target (this is an IIR filter on the value). It gets to largely the same results, but my hope is the values become a bit more stable that way. Probably starting the measuring over again on the back of this :)

v0.4.9

Choose a tag to compare

@github-actions github-actions released this 22 Sep 12:31

This update only adjusts some array sizes, since it wasn't handling array boundaries nicely (there was array stuff counting to <= the array size, and these typically start at zero and go up to a number one less than the array size). I don't think this'll help the slight irregularity in measuring super-high scores, but I've got a user who was seeing crashing and this might help :)

v0.4.8

Choose a tag to compare

@github-actions github-actions released this 09 Sep 19:16

Same as 0.4.7 with one bugfix: JUCE string handling was using a 'compare' thing to leave the score number off if it was 100 (i.e. if the song wasn't racking up a score at all, I don't show a number). Except I noticed that if the score is 10, that gets blanked out too! So apparently for JUCE string handling, 'compare' of "10" and "100" are the same because they start with 10. So now I check to see if it's 10 and I leave it alone if so, and that's the only change in this bugfix :)

v0.4.7

Choose a tag to compare

@github-actions github-actions released this 09 Sep 13:00

OK, here's where things stand with 0.4.7. I'm starting my whole measuring thing over. AGAIN.

It's partly because I have some subtle tweaks to how it measures at all: I struggled to get results more repeatable, and found that what I'm doing is so sensitive to conditions that I largely can't. Instead, it's a visualizer… with Opinions. I've explained a lot of that but also running all these experiments is in search of what Meter DOES want out of sound, and that's still being investigated. I know what I've wanted it to do, but I have to observe what it really does, especially when it gets sensitive and chaotic and measures stuff nobody else is beginning to deal with (like this 'cloud density' stuff it does).

I've rearranged how it does the 'bonus percent thing' it was doing, so now it's continuing to do that measurement of 'what proportion of the song is spent with the categories balanced', but rather than a perfect match or one letter off it's allowing two letters of difference among the meters. And instead of calling it a 'percentage of time in balance', turns out the evaluation of that is very prone to being twitchy, so I've flipped it: one percent of the time in balance becomes just the number '99', and ninety percent of the time in balance gets you a '10' score. It's like a top 100 on top of the existing ratings: if you can get it happening, the number will be counting up to number 1 and if the balance slips, it'll be counting down again. Seems fun if you're engaging with Meter on that level.

I'm still interested in what it tells me, I'm just making allowances for how experimental this all is. So now, when Meter really likes your stuff, it gives a number score, with number 1 being the very best you can be (and unreachable). Hope it's fun to play with. I sure am!

v0.4.6

Choose a tag to compare

@github-actions github-actions released this 08 Sep 18:52

This version adds to 0.4.5's 'plus' mechanic in a rather weird, maddening but fascinating way.

Where previously, Meter gave a '+' in front if the category letters were in perfect balance at the end of the track, I've added another mechanic, sort of gamifying it if you want to look at it that way. Now Meter tracks the proportion of time the track SPENDS in '+' mode, and if there's more than one percent, it shows you that as another sort of score.

The theory here is, much as Led Zeppelin intentionally tried to do with Stairway to Heaven, if at EVERY MOMENT during your track everything is perfectly balanced for that intensity level, and every time the intensity ramps up it's balanced across every aspect of sound, in theory that's amazingly compelling. It'd be like ramping from 'Sentimental Lady' to 'Born to be Wild' in one track.

I am damned if I can find any good examples of this, but I am still trying :D and all previous measurements etc. are as they were, so I don't have to start over remeasuring everything (got over 1000 tracks evaluated by now). The thing is, I've got several methods of metering going simultaneously now. The tone color should be very consistent and you're going for 'white balance', and it's directly shown by the first letter in the result. The plus (and percentage, now) is the far more twitchy subcategory grades, and my theory there is that it can represent energy ramping up over the track and pushing the outside of the envelope in each section… something that ought to balance.

For now… it's a new thing being done with the info that was already there, and if you were shooting for '+' and got robbed at the last moment, well now your efforts are properly documented :)