Questions recording in 32-bit float

[Glenn Stanton]

1 hour ago, RexRed said:

How do you know Melodyne (ARA2) is 24 bit? Please provide documentation.

there are some references floating around but it's easier just to see the file types it supports via the standalone app:

the ARA code doesn't seem to care about the audio depth as it's just coordinating the process - Melodyne itself uses the transfer and separation folders to store individual snippets (one good idea is to periodically it clean up).

[RexRed]

2 hours ago, Glenn Stanton said:

there are some references floating around but it's easier just to see the file types it supports via the standalone app:

the ARA code doesn't seem to care about the audio depth as it's just coordinating the process - Melodyne itself uses the transfer and separation folders to store individual snippets (one good idea is to periodically it clean up).

Thanks for the info Glen, Melodyne supports those files in that it will recognize them and then it "through the ARA2 specification" resamples these various file  bitrates to its own bitrate.

If it supported 32 bit float then Ulf from Celemony would not have said it does not support them. 

I am using 32 bit float files and Melodyne has no problem applying itself to the files but underneath the hood Melodyne has down sampled the file.

Where exactly does Melodyne store its files in a Cakewalk project?

The Melodyne folder in my current project folder is empty even though I have in on my vocal.

 

Edited August 1, 2023 by RexRed

[treesha]

My melodyne processed files go to c/user/ treesha/documents/celemony/separation And yes empty it from time to time. Not sure if they go somewhere else too. For me nothing is ever in the project file in the melodyne/ transfers folder. I think way back when clips etc had to be played to get transferred into melodyne they went in the project folder. 

[Glenn Stanton]

i  used to set up a dedicated Melodyne folder on my project drive for all separations -- but it looks like it changed in the not too distant past as i was checking the settings and found i could not change it! this is annoying since my documents folder is on my OneDrive which means if i forget to pause it, it will be conflicting with the synchronization functions. so i'll also need to exclude my cache folder from the AV as well.

then found in the help file:

Please note: The location of the audio cache is predetermined and cannot be altered:
Windows: C:\Users\USERNAME\Documents\Celemony\Separations.
macOS: /Users/USERNAME/Library/Caches/com.celemony.Melodyne/Separations

 

Edited August 2, 2023 by Glenn Stanton

[Starship Krupa]

On 7/29/2023 at 8:30 AM, RexRed said:

Perhaps, the biggest improvement of the VST3 plug-in is that it doesn't waste CPU resources and only works when it detects the presence of an audio signal, unlike VST2, which remains active at all times. For users, this means an opportunity to use a bigger number of plug-ins without overloading the system.

Based on how many times I see people repeating this piece of Steinberg ad copy, I'd say that it's one of the biggest misconceptions about the VST3 spec.

The VST3 spec doesn't require  or guarantee  that this feature will be implemented in any given VST3 plug-in, and so far, I've seen exactly one plug-in manufacturer implement this feature (Meldaproduction), and they also have the same feature in their VST2 versions.

The same much-repeated Steinberg blurb also trumpets that the VST3 spec now supports sidechaining, when I think we all know that sidechaining was possible for years with VST2's even though it wasn't part of the spec. I think the only plug-in maker who use the VST3-native sidechain feature are Waves.

The VST3 spec is just that, a specification. There's no magic hand that comes out of the sky and forces developers to implement all the features that are outlined in the spec. And since even now many developers still ship their products in both VST2 and VST3, the code used to build both versions doesn't implement the "advanced" VST3 features.

Since I actually have plug-ins that do the silence=sleep thing, I have to say that during mixing, it makes little difference. Plug-ins generally only eat resources when they're processing audio. Check Cakewalk's Performance meter. You can throw on 12 instances of iZotope Neutron and as long as you don't hit Play, the Performance meter will show that few resources will be eaten up. Hit Play, and that's when you better have a rocket sled computer, or dropouts happen. So the thing about it being something great that plug-ins go to sleep when they're not processing audio is kinda snake oil.

On 7/28/2023 at 7:08 PM, RexRed said:

I am wondering if singing my lines in key is not simply the best option.

Maybe @Byron Dickens could weigh in on this? 🤣

I will humbly suggest that if it's your own vocals that regularly "need" Melodyne, yes, learn how to sing in tune. It's not a mysterious innate talent (even I can do it). Your singing voice is an instrument, and the way to get better with any instrument is...practice (duh).

Put Cakewalk into loop record mode and sing a verse 5 times in a row over your instrumental track. Then play back each take. Listen for where you're not hitting the notes and take a mental note. Sing the bad notes a few times in tune to wear a groove in what being in tune sounds like.

Let your voice rest for a spell, drink some water, then repeat until you get diminishing returns. Do this for a few days (doesn't even have to be in a row) and you'll find substantial improvement. You'll get to the point where subsequent takes will be so on pitch that you can have a choir of Rexes all singing in unison.

Melodyne, IMO, is for fixing performances that you don't have control over, like tracks brought in by other people, and/or singers that want to sound professional without doing the work it takes to learn to sing in tune. It's much better if you don't have to fix anything in the first place.

As I said, I'm able to do it, and I've never had a voice lesson in my life. I just used the method I suggested (stumbled on it, actually) and it worked a treat.

Part of what you're doing is teaching your brain how to hear a note and then direct your voice to create that note.  Our brains love learning new ways to better control our bodies and will happily comply, working the problem even when you're not actually practicing. It's actually great fun.

Once you're comfortable with hitting the notes, your voice will relax and you'll get performances with better feel.

If this doesn't work, then re-record the lines where you missed the note(s) and comp. It takes much less time than farting around with Melodyne, and there's no concern about unwanted artifacts and such.

[Byron Dickens]

On 7/28/2023 at 9:08 PM, RexRed said:

I am wondering if signing my lines in key is not simply the best option.

Ya think?

[msmcleod]

8 hours ago, Starship Krupa said:

I will humbly suggest that if it's your own vocals that regularly "need" Melodyne, yes, learn how to sing in tune. It's not a mysterious innate talent (even I can do it). Your singing voice is an instrument, and the way to get better with any instrument is...practice (duh).

Just to add to this... practice here is really the key.

In my experience, there are two things that affect your tuning:

1.  Weak muscles, making it difficult for you to control/sustain your tuning.
2.  An untrained ear for tuning when singing.

I was very ill a few years ago and required a pretty aggressive course of steroids which weakened pretty much every muscle in my body (seriously there are muscles in places you didn't know you have).  My ability to sing in tune was pretty much destroyed.  I don't sing that much nowadays, so it is getting better over time slowly, but the more those muscles get used, the easier it is to hold my tuning.  Oh, and singing standing up is way easier than sitting down. 

Investing in a singing course, or at least doing some decent warm-up exercises is well worth it.  These work for me:

1. Start with the lowest note you can sing and gently hum 5 notes of a scale up and down.
2. Now start on the next lowest note and gently hum 5 notes of a scale up and down.
3. Repeat until you get to the highest note you can sing comfortably - don't strain to sing too high,  but do allow yourself to naturally transition into falsetto.
4. Repeat steps 1 - 3, using "N" sounds, then rolling "R" sounds (like you're doing a machine gun sound)
5. Finally, repeat steps 1-3 using "B" sounds - kind of like blowing a raspberry with your lips flapping.  Gently hold your teeth together, and put your forefinger and thumb around the sides of your mouth to stabilise your lips while they flap about.  The vibration of this final exercise on your vocal chords will massage them and slowly extend your upper range. It'll also relax your vocal chords making it way easier to sing afterwards.

As far as actually singing in tune in the first place, here's a tip - sing with a pitch correction plugin on and set the mix so you can here both the treated and untreated version of your voice.  Over a few weeks or months, you'll automatically correct yourself.  I noticed a similar phenomena when using the rasp effect on my TC Helicon VoicePrism: singing with the rasp effect on caused my voice to automatically produce the rasp itself without me thinking about it.

[Lord Tim]

Just to play devil's advocate here and give a different perspective, I have a couple of thoughts about Melodyne.

First, I think I do OK as a singer. I've been doing this stuff long enough by now that if I wasn't really good at it by now I probably should have quit to become a toilet seat polisher long ago (is that an actual job? Might be something for my retirement years... 🤔) but even then I still use Melodyne on nearly all of my tracks.

For backing vocal tracks I like to really lock in the pitch and timing to get it all super tight, and then decide how much looseness I want to leave in to make it still feel human. It was a constant bugbear for me in the early years when we had choir sections that weren't quite locked in and it still really irks me to this day! For our style of music anyway, some styles really lose a lot of the magic if you make stuff too perfect.

For leads, I always rely on the performance first and do multiple takes so I can comp the best parts to a good vocal track. Even then, I still use Melodyne as a safety net. I'd rather know that if I want to go for some outrageous note (and we're kind of known for our Stupid Vocalist Tricks) or some particular technique, I'm not having to choose between getting the right kind of feel or emotion into a take or absolutely nailing a note. You can tune a bum note, but you really can't fix a crappy performance, so having Melodyne there to tweak pitch and timing on an otherwise awesome take is really useful, and gives you a lot of confidence to say "f**k it" and just go for it.

What I've also found is, similar to what Mark was saying there, is do a reasonable take, tune the hell out of it, then sing along with it. The idea is not to keep that original ultra-tuned take, but it does help get you into the right zone for hitting notes for your actual take. I've had a few kind of green singers come through the studio here and that's helped them substantially.

The last thing is that maybe the style you're doing really requires crazy tuning on stuff where it's veering into the Uncanny Valley. It's not supposed to sound natural, but it's appropriate for the material. This is art after all, and is all very subjective.

But with that all said, I do absolutely agree with the sentiment that vocal training is a good idea. Whether you want to tune stuff in a crazy way or not, or use tuning as a safety net, having a solid basis in the beginning gives you options as to how you use these tools to create your art.

[RexRed]

I turned 60 in February of 2023, I started out singing in church at about 7 years old and started my first rock band at age 12.

I would have thrown in the towel long ago if I could not sing in tune.

Leonardo da Vinci was not a REAL artist because he used a straight edge to draw a line... lol

Singing a line properly not only requires timing, proper loudness and pitch but also one other element. Anyone know what that other element is?

And yes, I am classically trained as a vocalist.

Edited August 4, 2023 by RexRed

[Lord Tim]

I would say intent or delivery, which is the one thing I was getting at with my point - you can fix timing, loudness, pitch, etc. but you can't fix the intent, which is where something like Melodyne *won't* help, but it *will* give you a safety net if you want to make that your priority and fix any other shortcomings in the performance.

Please take no offence from my reply, by the way! That was more of a general comment than anything directed to you specifically  

[RexRed]

1 hour ago, Lord Tim said:

I would say intent or delivery, which is the one thing I was getting at with my point - you can fix timing, loudness, pitch, etc. but you can't fix the intent, which is where something like Melodyne *won't* help, but it *will* give you a safety net if you want to make that your priority and fix any other shortcomings in the performance.

Please take no offence from my reply, by the way! That was more of a general comment than anything directed to you specifically  

Your answer is very close and this goes not only for singing but other instruments as well.

I was attending college for music before I left high school. 

And I was reading college level textbooks for music as a youngster.

Though my knowledge of harmony is spotty, I took harmony and theory classes in college as well. ...and I had some GREAT teachers.

As well as singing in 500 voice choirs, I was also chosen to sing in the elite 50 voice choir that toured several states, and the prestigious 14 voice madrigal choir. That was at the university of Maine at Orono and I also attended the University of Maine at Machias. At Orono I was a performing major and all I had to do to achieve that status was sing for the professors.

Many years on the road singing in front of audiences in little towns and big cities getting home at 3am and I performed in these place live, just me and my guitar.

A lady once threw a diamond ring in my hat. She said, "I have to do it, it's from Switzerland". 

And one club wrote on the marquee, "The man who knows a thousand songs".

Now I have written and recorded near to a thousand songs as well.

I still own 2000 vinyl records, my collection used to be 4000 but I had to get rid of some. 

But, I can still learn new things, occasionally. That is why I am here.

On a break at a gig I once told three women that I had a love affair with a blank piece of paper and a pen. They all reached into their purses and gave me their fountain pens. And they were very nice ones.

My stories of the road are as long and winding as the road itself.

The answer is, "quality". And no, Melodyne cannot give you that.

So a line can be sung  perfectly in pitch, duration and loudness but as you sing that line over and over, quality can diminish with time. You have a small window to obtain all 4. 

Do you go with the quality line and fix the pitch or go with the pitch line? And, how much time do you have to devote to a single vocal track, or a line? When you need to, write the lyrics, compose the music, make the video, record yourself, fly the drone, play all of the other tracks, mix, master, publish, make the album art in 3d programs, Photoshop, build you own PC... and respond to hundreds of comments on your YouTube videos each month and make your own webpages etc...

Well, that is why were are called "artists" in this era of music democratization, because we alone make that call.

One of the greatest things I have learned is that people just want to have a good time...

Edited August 4, 2023 by RexRed

[John Vere]

I thought the word was going to be Talent. A music teacher told me once that the successful artists have at least 2 of 3 attributes. 

Talent, Ambition and Knowledge. The ones who have all 3 are lucky. 

At first I thought this was wrong but then you think of an artist you might  run into. Like the dude that will buy expensive gear and talk about gear forever,  and read all the magazines and can tell you just about anything about every hit record and who played on it.  They are always trying to put together a band and book gigs, but there's  just something missing when they play. Talent. 

Or the person who has lots of Talent and Knowledge but you can't get them off the couch, No ambition. You can go a long way with lots of Ambition and a little Talent. ( Taylor Swift?)  

I'm very similar to Tim in my approach to vocals. What I'm seeing for me is I just turned 70 so as I age the components needed to sing are wearing out. I defiantly cannot hear like I used to and my voice is not as smooth as when I was 50. I will re sing the same song I recorded 25 years ago and it's like a different person. But it is what it is, and I'm not trying to compete with anybody or score a record deal. I just want to enjoy the process without technology getting in my way. I'm real happy to be in a place with my set up that works without fuss. It is defiantly worth the time and effort to get the Hardware and software under your absolute control and leave that part of it behind you and start making music.   

I try and not overuse Melodyne and have slowly learned how to use the other tools it offers like aptitude and timing more than the pitching tools. So now I can just be loose when I sing and not obsess about perfecting the track.

Ultimately I like to get out in the real world and sing my songs live a dozen times or more before I commit to tape. Writing a song , recording a backing track and then singing the song cold will never work well for me. The singing has to take on a life of it's own, you might completely re phrase everything and choose a different melody even. It all happens when you get out there and belt it out in front of real people. 

And I'll end with saying , don't obsess about unimportant stuff like bit depth. It's not a deciding factor in making great recordings. 

 

Edited August 4, 2023 by JohnnyV

[Starship Krupa]

9 hours ago, RexRed said:

And yes, I am classically trained as a vocalist.

You have a truly astonishing background! Please forgive my musiciansplaining 😊. I went back and watched/listened to a couple of your CBBTV videos and your voice is GREAT. It gave me an opportunity to hear your vocals as recorded.

So I gotta question, even more now: why are you applying Melodyne to such a  beautiful instrument? Your raw takes I would consider "perfect" if they happened in my studio.

I have the same astonished feeling I get when beautiful women of my acquaintance mention that they want to get "work" done: why would you pay someone to take a knife to one of the prettiest faces I've seen?

Human voices do vary a little in pitch. IMO, at least, that tiny variation is something that our ears like. It makes a sound more interesting if it varies a little from "perfect." But in matters of taste, I suppose. You must be of the opinion that music like you want to make needs to have vocals that are as perfectly in tune as technique and technology can make them, and if that's the sound you're going for, I'll leave you to it.

One last suggestion: maybe you've hit the point where Melodyne is unable to improve upon what you already have.

[RexRed]

1 hour ago, Starship Krupa said:

You have a truly astonishing background! Please forgive my musiciansplaining 😊. I went back and watched/listened to a couple of your CBBTV videos and your voice is GREAT. It gave me an opportunity to hear your vocals as recorded.

So I gotta question, even more now: why are you applying Melodyne to such a  beautiful instrument? Your raw takes I would consider "perfect" if they happened in my studio.

I have the same astonished feeling I get when beautiful women of my acquaintance mention that they want to get "work" done: why would you pay someone to take a knife to one of the prettiest faces I've seen?

Human voices do vary a little in pitch. IMO, at least, that tiny variation is something that our ears like. It makes a sound more interesting if it varies a little from "perfect." But in matters of taste, I suppose. You must be of the opinion that music like you want to make needs to have vocals that are as perfectly in tune as technique and technology can make them, and if that's the sound you're going for, I'll leave you to it.

One last suggestion: maybe you've hit the point where Melodyne is unable to improve upon what you already have.

I guess what bothers me the most is that ARA2 and Melodyne are COMPLETELY silent as to what is really going on under the hood.

Google is silent about Melodyne if you search Google for Melodyne answers you will encounter my own Cakewalk forum thread inquiries which have gone mostly unanswered. And let's not forget that Celemony refuses to have a Melodyne public forum, you have to email them to ask questions and make requests. Thus, the answers are not public. This vexes me because I have a LOT of questions and requests. And getting an email back telling me I don't need 32bit recording or 96khz recording also concerns me. How low is Melodyne resampling my waves to? I am actually afraid to ask. 

It is like Bill Gates saying, no one will ever need more than 56k of RAM.    

Melodyne works with my 32 bit files but I am told by Ulf it does not work with 32 bit files. I am told that ARA2 and Melodyne are not resampling my files, so how is Melodyne working with my 32bit files if it is not resampling them down to a bitrate that it can recognize? And, what is that bitrate?

Why am I even recording in 24 bit if Melodyne is down sampling my voice to even 16bit?

I can hear a difference when Melodyne is placed on my tracks even if nothing is edited. There is an information vacuum online and Melodyne is the only real game in town. I adore Melodyne, while I am also concerned about how it is being developed. It does tons of stuff I don't need at all and the stuff I need seems unimportant to them, like colored blobs... I have been waiting for years for different colored track blobs.

As for "quality", when I analyze a line that is sung consecutively, it flows into other lines. When I punch in "like the old performers used to do". the line can sound disconnected from the previous and following lines. This is why fixing a simple bit of pitch deviation is better than re-singing the line . It is a "quality" issue where preserving the quality of the line is greater than re-singing it for a perfect pitch aspect and in turn losing the quality or, "one take" aspect.

But in order to fix that tiny bit of pitch deviation I need to place something on my file that is degrading the sample rate and "quality" of it. 

As I indicated, "quality" is one of the pillars of good music. Pitch, loudness, duration and quality... (You will find those four in any fundamentals of music text book. Yes, I read and studied them.) 

I like to know how things work. Just like I have my prepositions memorized, so I know how grammar works. I used to have all of my Euclidian assumptions, postulates, axioms, corollaries and theorems memorized for the same reason. And this is why I can excel at  3D vector art.

I don't like things that come in and "resample" my work without me knowing it. Because someone says I don't need higher depth or frequency files.

Or, I don't need colored blobs. Well it is hard to tell which blobs are from which tracks when they are all piled on top of each other.

Melodyne does pitch, loudness, duration and even formant  but it is the "quality" aspect that seems unimportant to its developers.

That quality issue is very evident to my ears. 

I feel like I am at a standstill with Melodyne. I am recognizing quality issues with the way it is working and I am told to "work around" those issues rather than Celemony addressing them and fixing them.

I mentioned variable bitrates.

If Melodyne threw more bits at the transients when it resampled the waves this problem could be solved. If a dialogue box popped up and asked us how we wanted our waves converted (like every other program in the industry does).

Melodyne could take these wave files and apply variable bitrates to the conversion process so the transients and sibilants would get more bits.

This is how video is done, when you have an explosion or detailed action sequence, you either raise the bitrates to an ungodly amount or you apply variable 2 pass bitrates. The computer searches through a scene and detects a part is complex and will throw more bits at that section.

While I am thinking that due to latency that Melodyne and maybe even Cakewalk are actually throwing less bits at the beginnings of sounds. I am constantly pointing fingers at what the cause of these quality issues are from. Thus the 32 bit interface, but the quality issues still persist in Melodyne.  

The proof is in the pudding and the transients, and sibilants and pitch bends are lacking terribly and completely un-editable in Melodyne. 

Some blobs you cannot touch in any way or they implode in quality.

And the answer to this problem should not be, "work around it".

If there is something I am missing here please enlighten me. I don't mind being wrong, that is how I learn.

Let's get this out in the open please.

Edited August 5, 2023 by RexRed

[RexRed]

Natural Vocal Pitch Editing Plug-in

https://www.synchroarts.com/products/repitch-standard/overview

Anyone heard of or tried this?

Edited August 6, 2023 by RexRed

[RexRed]

This is a great review of this device.

 

[RexRed]

This 32-Bit recording is EVERYTHING!

Today, (a few minutes ago) I turned off the padding switch on my mic and sang/recorded a vocal. This particular vocal I am singing Journey style vocals, very high pitched and very loud or "dynamic". You all recall having a song that you wanted to change your recording gain so the verse had a higher gain than the chorus... I am sure you do.  But when you do that then the verse sounds somewhat different in proximity/tone than the chorus. A huge dynamic range is necessary for recording a vocalist in their true epic form... So, in Cakewalk it clipped, WAY above the track envelope. I thought this vocal take is surely ruined! But, I just switched to clip gain, lowered the gain envelope on the clip and there was my vocal performance with no distortion whatsoever. I will NEVER go back to 24 bit recording. 

I might also add, when you record in 24bit when you get close to the clipping range you reach what I might "theoretically" call the saturation point. Where the file wants to clip and the 24 bit struggles with hearing the high end of the recording. This vocal I just sang was a mammoth of a vocal, and the 32 bit handled a tiger like it was a kitten.

When I think of other "mammoth vocals" I think of Mister Mister and the song, "Kyrie" a full-throated, belted out vocal. Bits do matter. 😊

Yes, you can turn your gain down for 24 bit recording, but who wants to amplify dirt?

Edited August 6, 2023 by RexRed

[RexRed]

I Just noticed that when you add Melodyne to a clip up in the play transport Cakewalk displays a progress bar and says 
"Mixing Down Audio".

[Quick Math]

Maybe you can test what it actually does to the file by using really pure sine wave, like recording 64bit wave file of a very pure (64bit) sine wave with a tool like BL sine, then load the file with ARA and do nothing and bounce it to 64 bit file and then check how much noise it adds by using SPAN with widest range setting.

And then you can also bounce the original 64bit sine wave file to 32bit and 24bit without any process and then compare them with the noise level.

I haven't installed Melodyne on my computer and also I'm away from my computer so I can't check it by myself right now.

I guess it is possible that Melodyne or the DAW secretly converts the file...

 

[RexRed]

19 hours ago, Quick Math said:

Maybe you can test what it actually does to the file by using really pure sine wave, like recording 64bit wave file of a very pure (64bit) sine wave with a tool like BL sine, then load the file with ARA and do nothing and bounce it to 64 bit file and then check how much noise it adds by using SPAN with widest range setting.

And then you can also bounce the original 64bit sine wave file to 32bit and 24bit without any process and then compare them with the noise level.

I haven't installed Melodyne on my computer and also I'm away from my computer so I can't check it by myself right now.

I guess it is possible that Melodyne or the DAW secretly converts the file...

 

I don't have the expertise to do this but you have given me a road map to possibly pursue this.

I would think this forum would have ready answers to this question but it seems there is silence in regard to Melodyne and what it actually is doing to our files.

I wonder if asking this question in another DAW forum might give a result.

I am sure Melodyne is not converting the waves to 32 bit and it is not touching the khs of the wave because Cakewalk is fussy about using the same khz across the project.

When Melodyne converts an entire track it is slow enough that it appears to be converting the wave simply to its preferred bitrate.

It would be interesting to open a 16 bit project and see if there is much conversion or if it is instant that might indicate which bitrate it uses natively.

I just did a test and when I added Melodye to a 5 minute audio wave it took the same amount of time converting the file to Melodyne as it did when I right clicked the file and bounced the Melodyne track to wave. So Melodyne is indeed converting the file and not just piecing it together. It is outright resampling the wave file.

I would think the answer to this would be of interest. And when you ask Melodyne to remove itself from your wave file, you do not get the same wave file back, you get the converted wave file. If you save previous project versions then that should keep your original wave files for retrieval if needed.

I might also add that a complete almost 4 minute song project with back up vocals, harmonies, ooos and ahhs, adlibs and tons of takes with 32 bit recording the project was around 421mb when looking at the properties of the audio folder.

Even 1 GB per song would not break the bank of my one of my 18TB drives.