Music Composers Unite!
Yes, I know I said I wasn't really interested in considering a DAW at this time. But given all the trouble I've had with Sibelius and NotePerformer, and given that the experts can't even be sure which software component is responsible, AND given that Avid doesn't offer tech support without an added cost contract, I figure I had better start thinking about a Plan B.
So... which DAWs are people using? What are the pros and cons of different DAWs? What is the workflow like when working in a DAW? Can you enter notes in something like the way you would in a notation software, or do you have to specify frequency in Hz and duration in seconds? Will I need additional hardware to be able to compose in a DAW? Are sample libraries compatible with any DAW or are they generally matched to a particular DAW?
I'm sure I'll have more questions as time goes on, but that will do it for starters I think... apologies if these questions are naive or poorly posed - I freely admit I know very little about the subject.
I investigated NP's website. I'm impressed with the ideas behind what they do. There were a few misleading statements made in the opening video. I found this to be disappointing.
The idea that typical well made sample libraries are one dimensional ( as stated in that video) just isn't true. While it is true that a single sample with no velocity programming or scripting is simply a sample played back, most sample libraries are varying degrees of mic placements with velocities programmed to key pressure or midi piano roll velocity programming. Not to mention round robins and other scripting tricks used to get realism in a sound.
Like two politicians who argue their worth based on what their competition isn't doing instead of what they can do. NP have chosen to focus mainly on a distorted view of sample libraries to sell their product.
It would have made the most sense to show more examples of what NP is capable of. They appear to be making great progress using their techniques but the initial sales pitch sort of turned me off.
Adding programmed attack transients to sounds in synthesis goes back before the 1980's.I could see where adding such an attack might add more realism to some sound samples.
From what you say about NP, they seem to be almost there in terms of getting their program to work well with notation. It's a shame they are so close to making the process seamless, yet there are still a few bugs. Maybe since you notified them of this they will find a way to correct it...hopefully.
I'm still getting used to the terminology used in digital audio, where words that are familiar to me in other contexts have very different - and sometimes unintuitive - meanings... velocity for example, which I gather means certainly not how quickly something is played nor even volume, but the force used by a human to generate a given dynamic. That's a good word too re: the distortion in NP, since it appears to occur only at high velocities and is completely unrelated to the volume of the sound: if you turn down the master volume in NP's mixer, it's still there at any dynamic level above mp. There is some intermittency to it though, as in some renderings it seems to be only barely audible.
But Wallander isn't aware of this even though I mentioned it. Arne calls the "swooshing" transient at the end of each note a "psycho-acoustic" effect, and he could be right. I am quite sure the distortion is not though, and he does not hear it. Some people (or more accurately, some ears) do, apparently, and some do not. Arne says that the timing issues will likely not be fixed until after things go back to normal after the pandemic, as they are short staffed right now. And in his last email he said that he was also hearing unevenness in the duration of some notes that he called "tempo issues", and he says that is in Sibelius not NP. The only tempo oddities I have heard is a tendency to rush some passages slightly, even with the Sib performance settings all at meccanico. An apparently experienced user on the Sibelius forum hinted that this is part of NP's "humanisation" - and thus a feature not a bug. It is a good thing when appropriate, but it isn't always appropriate, and it would be nice if it could be turned off.
It seems to me that with such a minimal issue in NP, it might be worth the wait to see if you can find a work around in it. You seem to like working in it. DAW's are going to be a learning curve. I find them interesting. Not everyone does.
I only recently realized that Jos Wylin uses Notion. This adds a lot of credibility to that program for me personally.
Well the NP issues seem minimal, but there is really no workaround to the timing issues with solo strings - except to go with string sections instead, a totally different sound. And then you run into the distortion. If you hear it, it isn't a minor issue. I tolerate it because at the moment I don't really have an alternative. Yes, I know DAWs are going to have a steep learning curve, and as I said, I've taken a pause on checking them out until I have my current piece finished. One thing is for sure: in making your score sound as if it's being played by real musicians, NP is simply wonderful, almost always. I wouldn't be able to do that in a DAW without a HUGE amount of work and practice, I'm quite sure. If you can get past the distortion, my Fugal Variations sounds amazingly good so far as rendered by NP, much better than I have any right to expect. I do like that about NP - but I'm very much aware of its flaws and they really do grate on the nerves.
When I think of distortion I associate the term with the electronics type. I understand there is probably lack of a better way to describe it at this point. If you posted an example of it I must have missed it. Distortion as an undesirable effect is when the input signal goes beyond the 0db threshold when measuring audio in db.. In the days of tube electronics the effect of going over 0db ever so slightly rounded off the peaks in a very desirable way and acted as a sort of warm compression. The reason why so many of the older mixing boards are so sought after, because certain mixing boards had a nice "sound" to them, and why to this day electric guitarists like tube amplifiers because they intentionally drive the tubes into distortion for a much more extreme sound that works for rock and metal music.
Later on when transistors and IC chips came along the circuit designs lacked that nice warm result. Going over 0db in that case results in a harsh annoying sound. Mostly a crackling noise. If a track goes into distortion the best option is to re track the audio. Ironically, there are now plugins that attempt to add that tube feel back into the electronics. I have a few of those. Many DAWs have mixing board emulation for this purpose.
The techniques used by NP look to be a hybrid process of sorts that uses samples along with synthetic attacks information.
No telling what the issue there is. My guess would be the samples themselves could have some issues in places. Maybe it isn't technically distortion. Could have been a recording technique they used added some undesirable after effects to a few of the samples or possible there is some issue with the way they are combining the synthesis with the samples?
The timing issues, if tied to the latency of the onboard computer audio could be the culprit. If the timing issues are more associated with a round robin type of articulation, IOW a pre programmed thing that's out of time, they should be able to re program it no problem.
I would be interested to hear the Fugal Variations when you complete it to see if I can pick up on the issues.
Have a Good Weekend!
I'll try to attach an mp3 of an NP rendering of an early fragment of the piece, the first 3 minutes or so (I was calling it a Ricercare back then). I can hear the "swooshing" effect as soon as the 1st violins enter at about 1:10, and the peaks of the expressive "swells" on each note from about 1:30 sound distorted to me. I have the NP master volume set to -4.0 dB, but I'm not sure what that level is referred to and have no idea if it keeps the peaks below 0.0 dB. However, monitoring the playback in Audacity, I never see the meters go anywhere near 0 dB, and the color bar indicators are always in the green. I think that it is probably, as you say, not true distortion but an artifact, either in the samples themselves, or of the way they are combining synthesis with the samples.
See the next thread in this forum for a discussion of the timing issues. I doubt they have anything to do with latency in the main board audio. They happen almost exclusively in legato passages: if two instruments are to attack simultaneously, NP handles that flawlessly every time. Arne Wallander says that the AI "humanisation" routines intentionally add random timing delays in legato playing to simulate what happens with human players in an orchestra section. Of course they shouldn't be doing that with solo strings, but he says that was an oversight and something they will eventually correct.
I'm not sure that's the entire issue though. As I mentioned, Arne also noticed some random variations of tempo in Sibelius that are probably contributing to the problem. And I have noticed that you only occasionally hear the problem in Gerd Prengel's renderings of his works using NP under Finale. So it doesn't appear that it is a single problem with a single solution. My workaround in my string quartet was to make several renderings, each with timing errors in different places, and combine the "good" passages in Audacity to make a single, flawless audio file. But that only works if there are enough silences that you can get error-free sections to splice together. If you splice at random points in the score, of course you get pops and clicks, sometimes quite loud ones, and Audacity usually can't do much to hide them. Fugal Variations has only a couple of pauses long enough to do this, and it would take forever to get error-free renderings of the music between them given how pervasive the problem is. That's the original reason I moved over from string quartet to chamber strings (4 players per section -- the timing errors mostly average out). As it is I think I might prefer this setting anyway.
Thanks again for your input and yes, have a great weekend!
This is a nice composition!
I listened for the issues you mentioned. The original samples recordings could have pushed 0db. When mixed down they lowered the levels. This would still retain minor clipping even at lower levels. After listening though, it doesn't sound like clipping to me.
I heard what sounded more like either some artifacts coming from the mp3. codec or what sounds like an exciter on those samples in the higher frequency ranges. The only way to eliminate the mp3 issue would be to hear the original wav file.
In one program I use called Ozone, I can drop a file into it, then demo it as a compressed file using different codecs. I can then opt to only hear the artifacts of the compression. It has a similar sound like high frequency whooshing or even a high whispering effect. This is pretty common in 128 or lower compressed mp3 files.
I am going to guess you also hear it in the original pre exported wav file. If this is the case, I'm going to say it sounds like they added some light exciter effect to the strings or maybe they bumped the 5K > frequencies a little too much there.. I didn't listen on my monitor speakers. Since I wasn't at my studio I listened on a pair of run of the mill ear buds. They seemed to pick it up though.
The attacks seemed more to my ears like a "breathing" effect. Not a deal breaker. I did notice it.
I agree multiple attacks on solo instruments are a no no. Maybe they can eventually make this a selected option in the software.
On editing clips- If Audacity has a cross fade option, this should eliminate the large pops you heard. I tried Audacity years ago but haven't needed it since. I though I remembered it has this option?
Best to you,
Thanks, I hope to have it finished in a few weeks - maybe 2 or 3. I am writing the concluding fugue now and will post a rendering of the end result to SoundCloud - it is already much too large to post here. I expect it will play a little over 20 minutes.
In Sibelius I hear the "distortion" during the original "live" playback too - it's not a compression artifact in the mp3. I don't know if codecs are needed for the playback itself - I assume so - but maybe it is happening in the playback processing within Sibelius. Whether it's a Sibelius issue or not I can't tell - my Finale trial has expired and I haven't heard anyone's string orchestra work rendered by NP under Finale. Yes ear buds pick it up just fine! I hate to admit it but the audio equipment I have for my computer is primitive - when composing I usually listen with ear buds. I have also burned CDs to see what it sounds like on a better audio system, and as expected it's the same.
What I don't understand is why the degree of the effect varies from playback to playback - sometimes it's barely noticeable, other times it sounds really bad. And it's not my ears, since if I export to a file, I hear it or not in that file - every time. And I always use the same mixer settings when I export to a file.
Audacity has a "click remover" and another tool that can suppress very short transients. I have found nothing else in their toolbox that helps at all with the pops. The click remover usually helps a little, but they're still noticeable. As I said, I may stick with chamber string orchestra for this work, just because.
A couple of thoughts Liz. The industry standard for removing noise from audio is Izotope RX-7. It's expensive but it will get rid of clicks and pops, maybe at some point you could get a free trial or find someone to use it to help you.
What your saying reminds me of ear damage. One of my ears responds painfully to certain sounds that don't bother anyone else. I've heard others complain about similar things. Tininitus is a different example of this phenomenon. Have you ever had an infection or other damage to your ear?. Just a possibility.
Interesting point about ear damage. Actually yes, my left ear had a sudden (temporary) hearing loss about 10 years ago and there was some distortion in my hearing in that ear for some time. The reason I don't think that's what is going on here is that I hear this ONLY in renderings made with NP. Not in MuseScore playbacks, not in Sibelius Sounds, and not in actual recorded string sound, ever, even at high volume levels. Also, if I turn up the volume high enough on an NP rendering, I can start to hear it in my right ear as well. The sudden hearing loss I had affected only my low frequency hearing - the high frequency sensitivity in that ear is still better than my right ear's, and I think my more sensitive left ear just picks this flaw up at lower volume levels. Also: some ears other than mine can hear it too.
Thanks for the suggestion of Izotope RX-7. Yes, Audacity is free and to an extent you get what you pay for. But I don't think I can justify buying a $300 package just for removing clicks that I shouldn't have to introduce in the first place. And I don't want to impose on someone frequently to do click removal for me. I'm constantly making small revisions to my scores, and I don't want to have to ask someone to help me out again. I'm going to hope that the timing issues can be resolved eventually (this problem everyone can hear and Wallander is aware of it) so that I don't have to do that kind of fine audio editing to create a good mock-up of my string quartet works. In the meantime, for the Fugal Variations, I think I'm going to stick with string sections.
I'm always suspicious of these blanket "noise removal" programs. They can't get it right every time. Removing a noise band is likely to mean attenuating part of the music signal as well - some will be embedded in the noise which means throwing it's balance with the rest out somewhat. It can only be some kind of notch or comb filter. Usually it depends on sampling a small amount of noise for analysis although I suppose it's possible to evaluate a noise band without that sampling. Does Izotope work like that?
Dealing with clicks and pops may mean reducing transients in percussion instruments.
I've used Audacity to remove noise but it's hugely time-consuming, zooming in to the audio sampling level. IIRC pops and crackles are in-phase so in a stereo signal they can be spotted. (Even if I've got that the wrong way around they're still easy to spot) then you adjust the level of the samples dot by dot. An LP can take forever! (That's what's attractive about balanced circuits - any extraneous noise cancels by being 180° out of phase.)
But it's a long while since I've done anything like that. Sometimes library samples have slight background noises (e.g. the Vsl strings harmonics - a very slight grating sound probably from being close-miked) and it takes time to work out the filtering.
All grist to the mill, I suppose.
By the way, Liz, the idea of velocity = amplitude came from early work on digital pianos when it was worked out that the loudness of a note depended on the speed at which the key was struck, and nothing else. The term caught on! (All these pianists who believe (and taught) that hand and arm contortions would squeeze and extra bit of tone from the notes were wrong. As Satie declared, he could play a note with the tip of his favourite umbrella and the tone would be same as played with his fingers!
I haven't used Izotope RX and I don't really know much about it but from what I've seen it's a little bit like Photoshop.
You want to get rid of that pesky in-law that shows up in all the family photos. You could just crop and take out half the family or just scrub out a spot and have a large blank area. But PS lets you fill in the cut area with appropriate background clipped from other areas of the photo.
RX has several tools some of which take the traditional approach of blanking certain frequencies which has obvious drawbacks. But the most advanced tool gives a visual representation of a recording with all of the sounds as shapes.
So theoretically you just snip out your problem and delete it and apparently you can fill in with other sounds to fill a gap, kind of like Photoshop. I can't recommend this but it is widely used in the industry..