Scoring Central

Here's a waltz that I composed over the last week as a sound test after giving my orchestra template an overhaul.

Pumpernickel and Periwinkle

Please rip it to shreds with any feedback about the composition or audio production. Thanks!

What you mean rip it to shreds. I wish I had 1% of your talent. It sounds nice and bouncy. I expected Mary Poppins to fly in my window at any moment.

(03-12-2018, 02:07 AM)bigcat1969 Wrote: [ -> ]I expected Mary Poppins to fly in my window at any moment.

Aww, thanks! I'm glad you liked it. Ever since I was twelve years old I've dreamed of composing music that makes people feel emotions, it sounds like I was successful this time.

(03-12-2018, 02:07 AM)bigcat1969 Wrote: [ -> ]What you mean rip it to shreds.

Ha, it's mostly just that I feel like I'm still developing my ability and ear for putting together all of the pieces in a pleasing way: the melody, harmony, rhythm, and timbre of the arrangement has to be just right, and after that there's all of the audio production stuff like balancing the decibel levels of the different tracks, creating stage presence with stereo effects and reverb layers, along with lots of other potential audio effects that I don't really understand so I don't even use them... and of course I got to that point that I knew that I had twiddled around with everything enough and should just share the rendered audio track with people even though there's always just one... more... thing... that I want to try... So yes, any feedback would be really appreciated!

Nice composition and use of form!

I think the mix could use some improvements- it seems a bit too "small room"-y to me. Consider using a bit less and a big longer of an impulse/reverb tail, about 1-3 seconds is optimal, and don't be afraid to use a bit of predelay. My process is to start with the reverb all the way off (only the dry) and slowly increase until the space just starts to appear, or vice versa, start full reverb and pull back until it feels like you are the conductor standing in front of the ensemble. I only use about 20% to 80% wet-to-dry mix (or less), or with a decibel meter, 0 dB dry and maybe -12 to -20 dB wet. I cannot emphasize enough how important the final step of 'pulling back' the reverb until it 'sits' right is. I think snares are just about the best indicator of how well a reverb works, as they are basically a broad spectrum impulse of noise.

Many of the sounds (in particular the woodwinds and percussion) generally felt like they had some filter on top of them getting rid of all the highs, like they were recorded onto an old Type I cassette tape or an old sampler in mono from too far away, then panned around. In some realms, this is considered beneficial, but if your goal is to create a realistic orchestral sound, I think cleaner, closer, true stereo sounds would go a long way towards that goal.

My best advice would be to spend more time in real halls, rehearsal spaces, etc listening to the way the instruments actually sound in the space. There is nothing more beneficial to orchestral mixing than that, in my opinion. Once you have an idea on what makes a "good" sounding space rather than just a "space", you will be better able to find and spaces that bring out the best in your samples.

(03-12-2018, 11:42 PM)Samulis Wrote: [ -> ]I think the mix could use some improvements

Now that's what I mean by "rip it to shreds"! I really did try to turn down the reverb based on the last time you gave feedback on a composition of mine! One of these days I'm actually going to get the reverb right.

I'm using three reverb busses, the middle one (harp, piano, woodwinds, and basses) being a bit more wet than the specific numbers that you have recommended. I made the back row (percussion and brass) about 10% more wet, and the front slightly more dry.

I'm baffled by the recommendation to not be afraid of predelay...  My front row has 50ms, middle is 30ms, and back is 0ms. Would it help to increase the first two numbers?

I set the room size to 30 meters, and the mid frequency decay is 1.5 seconds, but the reverb is set to decay high frequencies much faster than that. Based on your comments it seems that I should let the high frequencies ring out a bit more than they do right now. There are no filters on the dry signal, but my reverb does have a low pass filter set at 7.5kHz for the woodwinds and 6.0kHz for the percussion, which was an attempt at avoiding a harsh reverb in the high frequencies. I wonder if it would help if I were to make the whole thing more dry like you suggested.

Finally, I'm using Virtual Playing Orchestra more or less as it comes, except that I narrowed the stereo width of each woodwind to 0%, since I'm only using the solo ones, it seemed like they shouldn't have any width at all. Is that possibly causing the cassette tape sound that you're talking about?

I'll tweak around with this some more, I feel like I'm finally getting better at the composing style that I want, but I'm still lost when it comes to the audio.

Edit: Here's a new attempt at mixing, 85% dry, 15% late reverb, increased predelay, and adjusted the high cut/crossover/mult parameters to let the upper frequencies ring out a bit more: https://instaud.io/1TKH

Michael, why don't you upload the three stems (or even better, the raw MIDI) and I can show you how I would approach it? I think you are perhaps over-thinking reverb when instead I think it is better to simply learn how to hear it. There are no good general rules for reverb- each plugin is different, each space is different, each sound source is different. We could be here until the cows come home talking about this number and that value and looking at books on acoustics to find recipes for diffusion, but at the end of the day, all that matters is- does it sound good or does it sound bad? Does the reverb sound like a concert hall or does it sound like the living room?

I would really advise against squishing anything down to 0% width unless you know it was recorded in a coincident (XY, Blumlein, Mid-Side) array. The common recording methods for orchestra (Spaced Pair/AB, Decca, ORTF, NOS) will all cause some undesirable phasing when squished 100%, which can cause a filter-like effect. Even then, any sample which already has reverb/"Room" sound to it will not fare well if squished to mono, as the cues of that "room", all those little reflections and resonances, will be lost and turn directly into a kind of audio mud in which the ears of the listener get readily stuck.

In essence, a mono signal suggests an instrument projects sound in a straight path, like a laser beam. However, most instruments, especially woodwinds, emit sound in a diffused nature, with the sound going off into the room in all directions, like a candle or a light bulb. That light bounces off of any and every surface, before reaching your eyes. While it is possible to convert the light of either of those into a kind of "beam", or with a great deal of knowledge, even possibly into a laser, the result is quite a bit less suitable for, say, illuminating a space. Try sticking a mic into the bell of a clarinet vs. putting it a foot off the middle of the body. The bell is almost just there for decoration when your fingers are only on the top joint! The clarinet, and all other woodwinds and almost all acoustic instruments in existence, are non-directional, thus they should never be condensed to mono if it can be helped.

Compare these three examples-
Original Stereo: https://instaud.io/1TPI
50/50 Mix Squished to 0% Width: https://instaud.io/1TPJ
Previous with reverb put on top: https://instaud.io/1TPL

There's only one decent way (that-I-know-of) to re-inflate a mono recording and that is through careful use of convolutions, a little stereo spreader, and a lot of patience, and it doesn't even work all that well. See Eduardo's Tarilonte's libraries such as ERA (recorded in mono, inflated with convo's)- there is an unmistakable stereo field to those sounds that is just "off" from what we are all used to hearing that makes them very hard to blend with "normal" stereo sample libraries.

If you want to work from mono recordings, work from very isolated/close ones, and use some sort of room modelling plugin or lots of convolutions. Any sort of room sound in the original recordings will destroy everything you try to put on top of it, especially with snares, which as I said, are the absolute pickiest of all.

(03-13-2018, 05:05 PM)jmcmillan Wrote: [ -> ]Good stuff! I liked the melody and the orchestration!
Some composition critiques (small stuff!):
1. I would delay the suspended cymbal rise one beat so it peaks as your going into the new phrase at 00:40. Felt too early to me. 
2. The accompaniment from 00:40 to 00:50 changes from 1-2-3 to 1-2-rest and 1-rest-rest. Felt “studdery” to me. Maybe keeps it 1-2-3 all the way through or use 1-rest-3?

Thanks J!

I was experimenting with timing on the percussion, and specifically with the suspended cymbal I tried getting it to parallel the last four notes of the piano solo (The root, 3rd, 5th, octave part). I didn't think of making it peak on the downbeat like you suggest, I'll have to try it.

The changes in accompaniment were my attempt at prevent the same "oom bah bah" thing going on for the whole piece, I wanted it to have some variation. You might be onto something with that 1-rest-3 idea though, that might be enough change without making it stutter like you pointed out.

(03-13-2018, 03:30 PM)Samulis Wrote: [ -> ]Michael, why don't you upload the three stems

Here you go, I widened the stereo width of the solo parts.

Alright, I played around with the stems for a bit. I have a few general suggestions which may help in this particular use-case-

1. It's best to avoid mixing instruments of different amounts of baked reverb/room into the same group, instead separate out instruments of abnormal wetness/dryness to their layer into their own groups (the super-close mic'd solo trumpet vs. the percussion, the overly wet upper strings vs. the super dry bass, the damp flute vs. the dry piano).
2. That snare is really low-mid-full. I would cut out most of the mids and boost the highs a ton on it so it has that 'sizzle' or otherwise find a different snare. It's also kind of 'damp' and very centered, so it really hates any and all reverb with a passion...
3. In general, percussive sounds can 'take' less reverb than sustaining sounds. Strings and woodwinds can soak up tons of reverb and sound great, but a little too much on a snare and it sounds like a dish disposal eating a wooden spoon.
4. The best samples for adjusting are recorded neither close nor far but at a comfortable distance.

Basically I just made an eq which looks like a big crater and tossed a ~3s concert hall convo reverb on the output, then a bit of compression and an exciter to get a little more sizzle.

Even with that, I didn't have much more luck than you to be honest. It's very tricky to combine these sounds well, it really needs to be mixed per-sound, possibly with some processing of the samples beforehand to try to mitigate the baked reverb. It's sort of like having a bucket of wet cement with chunks of dry in it then pouring water on it, versus a bucket of dry cement and pouring water on that. One bucket is going to be a sloppy mess no matter what you do with it until you manage to get it dry enough to work with and those pockets of dry cement are going to cause problems unless you break them down and mix them in proper, the other is going to be good to go so long as you don't under or over add the water. If you already have your bucket already a little wet all throughout and you just add a tiny bit more to get it to the right consistency, that's ideal.
https://instaud.io/1U7t

Here's a piece I remembered I did years ago using a bunch of free soundfonts. Each instrument had at least one EQ and faux stereo effect on it just to get them to work in the space and I had to do a lot of digging just to find sounds that would mesh right. It's still a bit of a mess after all that-
https://www.newgrounds.com/audio/listen/568889

The kind of sound which is most flexible is a mid-range stereo recording, with little room, but far enough from the instrument to make the proximity effect negligible (about 3-8 feet, depending on the space)-
https://instaud.io/private/33ccfb6d53080...4a8709155e

Now I will magically transport the viola onto the stage, as if we were listening from omni mics at about 10 feet-
https://instaud.io/private/628b7384c68a7...89a10a9031

Now we will go back to off the edge of the stage-
https://instaud.io/private/3b112d7e4e303...6195b46c98
(I applied a little stereo thinning to about 50% and stereo pan slightly left to the concert position of the viola, then attenuated the highs and lows a little, then put on some convolution reverb)

It's important to keep in mind when using algo reverb rather than convolution that in a convolution is a "baked sense" of the absorption and diffusion of the space, something which must be modeled in an algo. Absorption depends on the materials in the space and their exact location. Although it may not be noticeable, moving a single baffle on a stage does adjust the acoustics of the space. Typically absorption eats up your high frequencies, but if you're recording with cardioids, you will lose quite a bit of lows as well. Besides, cleaning out some of the mids (250-750) is almost always a good idea before sending something into reverb, as that is where mud often is the nastiest (although low mud as well is not nice). That is a bit of what I have simulated here. By recording far enough away, I managed to avoid the proximity effect and the undeveloped tone of the instrument, but not so far that the existing ambiance would make it difficult for me to change the space. Rather, as you can hear (and are welcome to try for yourself on the original dry track), it takes to the reverb very nicely and is easy to move around between spaces. I can even put it in a different sounding studio if I wanted to-
https://instaud.io/private/4efa179b34082...8a98bdcace

Edit #5000: In short, when 'shopping' for samples, look for things which when dry resemble that, so that you can move them and work with them as easy as possible.

I definitely agree with Samulis here. The differing wetness of the samples used is problematic, and basically everything would need to be processed separately to make it all gel better.

Here's my own quick take on it. I tried moving some stuff back a bit in case that would help, but you can clearly hear that some instruments have sort of a "double ambience" to the point where they sound like you're listening to the 100% wet signal. A nice track though! Very memorable and hummable.