I just did this experiment .

record a singing part in 32 bits, 24 bit and 16 bits wav. I raised the fahters to +10, both for the track and the master (the distortion was very evident)and then export them in different depths 32 to 32...32 to 24...etc.


and i got this when i reopen the files and normalize them

1. 32 to 32 bits does not distort when normalizing
2. 32 to 24 bits distorts when normalizing
3. 32 to 16 bits distorts when normalizing ( if it distorts in 24 sure in 16 too but still tried)

1. 24 to 32 bits does not distort when normalizing
2. 24 to 24 distorts when normalizing
3. 24 to 16 ( sure is the same like 24 bits)

the test in 32 bits, since it would be obvious the other 16/24 would distort

1. 16 to 32 bits does not distort when normalizing


Well now my question was what would happen if I open a 24-bit flac file and exported it to 32-bits ( Wav)

1. 24 bits Flac to 32 bits Wav does not distort when normalizing

I did not know that 32 bits float had this property, I say, of maintaining the dynamics, since when normalizing the file, there is no distortion and it recovers the dynamics, I did not analyze if it recovers 100% but to my ear and seeing the waveform I would say which is the same. So I think that if one is going to export a track, guitar, bass, synth, whatever, it would always be ideal to do it in 32 bits

 

SOS Technical Editor Hugh Robjohns replies: I’m not surprised you’ve had a problem with getting 32-bit floating-point files to play in Serato, or any other conventional playout system, and in fact the Serato web site lists only the familiar standard file formats as being compatible — ie. WAV and AIF for conventional fixed-point formats (up to 24 bits), ALAC and FLAC as lossless data-reduced formats, and MP3, OGG, and AAC for lossy data-reduced files. So yes, Serato would have been fine with fixed-point 24-bit WAV files, but wouldn’t know what to do with 32-bit floating-point files at all.

32-bit floating-point (and other similar large word-length) formats are intended purely for the internal signal processing of a DAW during mixing, and are designed essentially to maintain the quality of individual sound elements at wildly varying levels while performing complex signal processing. Floating-point formats aren't intended, appropriate or efficient for use as a delivery medium for consumer applications.

So, by all means, work on and store ongoing mix projects in a 32-bit floating-point (or higher) format — most DAWs do that anyway by default. But for taking finished mixes to a mastering house (or for sending out as auditioning files) a fixed 24-bit word-length WAV file is the most universally acceptable format. Ideally, you should retain a few decibels of headroom, too (ideally 3-6dB) — don’t normalise it to 0dBFS because that will make life harder for your mastering engineer!

As always, there are caveats, and one is that if you’re expecting your mastering house to re-balance individual stems as part of the mastering process, it might be more practical to send the whole DAW project still in its native 32-bit floating-point file format. However, there might still be a potential problem of incompatibility because there are many different and incompatible 32-bit floating-point formats employed by different DAWs, so check that the mastering house can deal with your specific DAW project format first!

In general, though, if you’re sending straightforward stereo mixes to the mastering house they will be expecting standard 24-bit WAV files. So, I’d recommend keeping your work-in-progress projects in the DAW’s native 32-bit floating-point format, but bounce mixes out for auditioning or for sending to a mastering house as fixed-point 24-bit WAV files — simply because you can guarantee they will always work with all standard external audio-file players, DAWs, and mastering systems.

www.soundonsound.com/sound-advice/q-32-bit-floating-point-or-24-bit-fixed-point-audio-files

 

This piques my interest. It's true that decompressing FLAC is not big on the CPU. Still, a DAW sometimes can need every bit of CPU.
Are we sure internally the DAW doesn't convert those files to wav or whatever format for performance reasons?
I mean, even the waveform views (peaks) are cached in the disk with some internal graphic format. It's far from an apples to apples comparison but I wonder.

 

Last edited: May 3, 2022

Before converting the flac files (Not all, some because this is still in doubt). I told myself what happens when you reopen it in the Daw, and if you save it as "Wav" it has exactly the same weight as the file had before being converted to FLAC. Apparently the only thing it does is compress it and then decrime it

PS: Nice to see you again!

 

Most DAWs do exactly this. Compressed files (FLAC, MP3, AAC, etc.) are decompressed upon import. Some (I believe Reaper for example) can decompress bigger files in chunks (i.e. during playback) to save RAM. But even this shouldn't be a problem with today's CPUs. Decoding is a fairly simple process.

 

Hasn't this sort of thing just become personal preferance now?

Yes you can save space by changing your wave files to Flac but is the space you create really worth the amount time you spent for the money you save not buying the bigger Hard drive?

If I remember the principles behind Flac, it's like a zip file, it uses a deflation method via standard bit patterns. The more compressed the file is, the less space it takes and the more processing power is needed to both create and uncompress to play it.
Most software only have 1 option and that isn't very processor heavy and dosen't reduce the file size much. A ratio of around 3:4 some times 1:2.

I don't understand why Ryck is having distortion when normalising. if your normalising peak to 0dB then no matter the file format you should never get distortion. you may have nosie generated at low =dB levels but there should be no distortion. If your getting distortion when normalising peak to 0dB then you should check your codec, somethings not working right.

If your Noormallising with your peak >0dB then you'll get distortion with what ever format you use, it clipping.

if your chaging a 32 bit file to 24 bit, your reducing the file size by 1/3 without changing the file format.
So do you really need to convert to Flac if your doing this?

 

The meaning of "Flac supported natively" in the specs of the DAW means that there is no transcode: file will be played as it is.

This idea of compressing any wav to flac is interesting, it sure can save space, but imagine that moment after transcoding everything in flac, when you have to press "delete" on all your wav collection of masters & samples created during all your life, that would be really painful... no way, I'll stick to the wav

 

That makes sense, It still means the file the DAW uses is uncompressed. Nice... again? No effing clue...

That makes sense. But my question was more if they save the decompressed versions in some sort of project file, cache and stuff. It's what I'd do. Reaper allows you to set this option if I recall correctly when you save the project. But I wonder what the rest of DAWs do by default

As long you're going to use it for editing/DAW it's better to have it in WAV. Converting to FLAC is more for your music for listening collection if you don't use MP3, for instance. Or for a finished project's final render.

And then there's this one-of-a-kind Plugin Boutique Scaler's FLAC use for its library of one fuckin' Giga
Don't know why so huge. Well, I slimmed them down to 16 bit 24K just because I don't need them more than for perhaps a quick test. Kontakt style but hardcore

 

Last edited: May 4, 2022

When I talked about distortion, it was when I did the test of raising the levels of the track to +10 , to distort on purpose, and see if really 32 bits float kept the dynamics when exporting it and reopening it. But if a track exceeds +0db and you distort and then save it in 24 or 16 bits, when you open it again, it's already like that, distorted, unlike a 32-bit floar file that you can recover its dynamics

I need to make space on my drives because they are almost full of so many libraries and samples. And now I really can't buy a hard drive, of course it's the best solution.
and I also don't want to sacrifice the kontakt libraries. I have already sacrificed several. It's all the fault of the sister site that makes me fall in love with its kontakt libraries hahaha

 

It is best to store some rarely used Kontact memory hogs:

Sign up for MEGA now and get 20 GB of online storage - for FREE
https://mega.nz/start

 

Do this test:
convert an original.wav to original.flac;
then reconvert it to reconverted.wav;
use some comparing tool, if original.wav is equal to reconverted.wav;
repeat test as many times you can to assure it's a safe procedure, if yes, delete wav files.
After all, if flac is lossless, that should be the result.

 

It is losslesss. It's a fact. The only thing that could alter the sound if you convert a format FLAC doesn't support.
Someone mentioned one before.

There's an ultimate and very funny video debunking some crazy ignorant claims that FLAC sound "worse" than WAV. Pseudo-snob-elitism at its worst. I had a lot of laughs with it. Priceless.

 

Last edited: May 4, 2022

Wow man, I took a look at the WavPack format features and it's crazy impressive.
FLAC is fine and perhaps the only con is not having float support but it wasn't designed for that.

But man, WavPack has some mind-blowing possibilities. I just had a techie boner. Sorry, that was rude. I meant techie "true peak"

 

FLAC is more supported than WavPack or ALAC
Like MP3 is more supported than OGG or AAC

"Better format" is only when you CAN use it in the real world

PS : i maintain my point on 32bits : it is USELESS as storage format
Like 192 kHz
24bits 96 kHz is more than enough for anything ... until you mix for bats

 

You're probably mixing up file (or 'audio') formats and dynamic range. 32-bit float has a dynamic range of 23 bit* with every fixed gain. So it's less than 24 bit.

*Yes, it can use the sign to add 1 bit and get 24 bit, but not more than that.

 

Last edited: May 12, 2022

The thing with precision is that when you know the absolute maximum and minimum, the case with waveforms, you have it more easy to convert between 24 and 32 float.
I don't remember exactly, but I think you can do a 1 to 1 conversion each way.

 

Last edited: May 13, 2022

you say that a 32 bits float file is actually 23 bits?

 

So far I have worked with some loops in Flac and I have not found any difference with Wav. I've stretched the loops, changed the pitch, saved the file, reopened it. I find no audible difference really.
In my opinion, the only difference between Flac and Wav, is that the Daw will have to unpack the Flac file and load it into ram (I think) or host that decompression somewhere.

Where I did find an advantage is in 32-bit float wav files, because even if you go over the 0db peak, your file will not distort and you can recover the original audio. This flac can not do because it is not 32 bits. That is to say that if you save an audio in flac and it goes over 0db, those peaks are not recovered.

 

32-bit float is IEEE 754, so there's 23 bit mantissa + 8 bit exponent + 1 bit sign. So you basically have a dynamic range window of 23 bits (or 24 bits with the sign), that you can move up and down in the domain. Hence, the dynamic range cannot be higher than 24 bits for any given gain level. It's also the reason why there is no 24-bit float (or even 16-bit float) in audio production. The dynamic range would be too low.

That's exactly the point of the float format. Gain staging between the A/D and D/A converters becomes negligible. It's useless for finished mixes and masters, but really useful for all the processing before that.