Why you should mix down at 32 bit float!

[@tomaszmazurek64]

I will try to explain the math behind this without it being too techy and with just a little bit of lying. 24 bits is roughly equivalent to 7 decimal digits. So an equivalent decimal format could store whole numbers from -9,999,999 through 0 to 9,999,999. A 32 bit floating point format incidentally also uses 24 bits of significant digits and 8 bits for the position of the point. So our hypothetical decimal floating point format could also store numbers from -9,999,999 through 0 to 9,999,999 and specify that the point is behind the last 9. If the DAW tries to export a sample value of, say, 6,000,000 both formats will store it equally well. But if the sample value is larger than 9,999,999, for example 22,111,234 the first format cannot store it very well - the best it can do is store the largest number it knows - 9,999,999 and so we get clipping. The floating point on the other hand has a trick up it sleeve - it can store 2,211,123, a number ten times smaller, and specify that the point is one position behind the 3 - giving us 22,111,230. As you can see we lost the last digit of precision, the 4, but the stored value is much closer to the original and the loss of that last digit will be largely unnoticeable. The binary formats work quite similarly with the difference being that they use binary digits (0 and 1) and powers of 2 instead of 10 and a few implementation details, like the way the sign is represented or a small trick the float format uses to store 24 significant bits as 23.

[@mxiii4980]

you made me feel stupid

[@zerix01]

This actually helped me out a lot.

[@gatsnguts]

One of the best explanations yet!

[@emiel333]

That’s the magic of 32-bit single-precision floating point operations. 16-bit is half-precision and 24-bit somewhere in between. The lower the bit rate, compared to 32-bit single precision, the more “mistakes” are written to the file. It’s less precise. When I use Logic, I sometimes set the audio engine at 64-bit double precision, it isn’t necessary, just like you explained, 32-bit FP has the magic trick already in its skill set. Imagine how big the file size would be for an exported orchestral music piece in 64-bit at a minimum of 192kHz or higher… Inside the digital audio workstation, the dynamic range is unbelievably high when working in 32-bit. I can’t come up with the exact number but remembering it is somewhere between 800 dB and 1000 dB of dynamic range to work with. It’s such a high value that it isn’t necessary (24-bit already has a dynamic range of 144 dB when inside the DAW) for dynamic reasons, but more for the occurrence of clipping. Better said, the lack of clipping. Although a producer shouldn’t let their channels and the master peak above zero. You won’t here it until you have exported the music file. So stay below zero everyone. On channels and especially on the master (or 2bus) out.

[@emiel333]

Make sure when mastering a 32bit mix, that it isn’t “digitally” clipped in the DAW before you export it. If the master channel goes above 0dB LUFS, you won’t notice it while inside your DAW but you do hear it when it’s being exported to a wave file. So make sure all channels stay comfortable below zero and leave a headroom of -4 to -6dB so the mastering engineer can make adjustments. Also, if a 32-bit (single precision floating point operations measured per second in FLOPS) is exported well, if you master it yourself and must convert to 24 or even 16-bit don’t forget to use dither.

[@Rolanoid]

Ever since I started adding analogue emulation plugins to my projects I'm still careful not to unintentionally clip these plugins even working at 32 bit. I now try to get in the habit of -10db peak or -18 db RMS to give everything including the master buss enough headroom. Then in mastering I will raise the level as needed. My mixes sound better for it.

[@luisfarias134]

Yes, lowering all track volumes down and turning it up in the final stages is the way to go!

[@andrewsears3030]

Your excitement about this tech warms my soul.

[@toddsmithmusic3172]

Whats up gaz !! HUGE FAN !! I love your passion for electronic gear man, it just drops from you when you get near gear. I also wanted to state I love how open minded you are about new gear you've never touched, so many synth players draw lines in the sand where you feel like a open book. It's refreshing. I've been making Volca based jams which got pretty popular and have grown into something more than just "jams" so I've been looking into mastering final products, this info helps a lot thank you !!

[@fuzzupuzzu]

Nice demonstration, ive been using it since long but good to actually see the literal effects of it

[@roelasedillas7912]

this is the technique that the main stream recording studio don't want us to know, thank you!! you save my projects!!!!

[@SoundAuthor]

Thanks for this Gaz. I've always wondered what the whole "floating point" thing was all about. Even though I've only got an abstract idea of what's going there, I at least have some basic understand of what it can be used for now.

[@elijahmant2855]

Gaz. Good on yah. Very well and to the point explanation. Appreciated!

[@TozBourne]

Great stuff, and...first in a series? Oh yes, please! I look forward to more installments.

[@danvercillo]

This is great Gaz, learning from you is quite awesome.

[@smartti1970]

can we hear more of that track it sounds gorgeous !

[@SoundsAndGear]

Great video man!

[@jasabasenara8124]

Thanks for that. I never knew it wouldn't clip at 32 bit float.

[@JohnnyNatrium]

Well I feel like the point is that, it will still clip if you load up the mix with rendered audio again and keep things too loud. But with 32FP or 64FP, you won't have actually lost the information from the rendered mixdown sound because that is still stored and so you can always go back and turn it down and all of the quality is still there. It means you can mess around with the audio a lot and if you add too much depth to the sample it can still be tweaked without loss of quality before you finally master it at 24/16 bits. (Edit): I had only watched half the video but yes it turns out he already points that out pretty clearly himself here so I'll just say I've reiterated it just to make it clear since it technically isn't correct that the mix won't be clipping just by rendering it at that depth.

[@jasabasenara8124]

@@JohnnyNatrium cool. Thank you for the explanation 👍

[@dirtyb82]

Good stuff , thanks Gaz .. Look forward to more !! Ace !!

[@RandomNeeds]

Nice quick explanation Gaz.

[@Psychlist1972]

That's some pretty fantastic info there, Gaz.

[@nnxmediagroup5049]

Very clever Gaz. Thanks

[@EdExploresScotland]

Excellent explanation. Thank you.

[@andy-ford]

Cheers Gaz that's a very handy tip!

[@bdw0lf]

Awesome :D Really excited about this new series! Thanks Gaz.

[@KevinJanosky]

Beautiful. Well done sir.

[@Padbangers]

Great explanation Gaz! Thanks for sharing

[@VacancyOfDisco]

Yes! More more more! I love simple studio tricks like this

[@strangerhertz5687]

Nice instructions boss! What if you pre-master directly in ableton, ya know adding some limiters on the master channel. Should you still do 32?

[@GazWilliams]

Best not to limit a premaster really as that affects the scope of the mastering. Using a mix limiter can be helpful but turn it off when you make your premaster mix down.

[@strangerhertz5687]

+Gaz Williams thank you sir!

[@twzstudio139]

A great demo, and I agree that using 32bit makes total sense. The reason this works is because bit depth effects dynamic range / headroom. It is still possible to clip at 32bit, but you would have add insane amounts of gain to do so. Do you think tracking at 32bit makes sense as well, or simply saving that for the stereo mix only?

[@annother3350]

@ but in 32-bit case it's a fucking mountain of garbage.

[@sysexstudio]

That extra 8 bits of dynamic range is quite a lot when exponentially lopped on top of the 24 bit range.

[@380stroker]

No one records in true 24 bit.

[@mikeysharp6280]

Gaz is so chill!!!

[@telogic1]

Just remember gaz that 32bit float uses 50% more hard drive space and if your mix is clipping when you mix down at 32 bit . When you adjust for clipping you are also reducing the overall mix .

Yes, this video can be very misleading. It's more important to record a clean signal to work with regardless of 'bits' used.

[@nekistayfruity]

wow i dont knew that thanks for that lesson

[@cogbeatz]

Thank you so much Gaz!

[@bencawood]

Are you recording in 32 bit too ?

[@pryere]

Thanks Gaz, very useful. Much appreciated.

[@manecolooper]

Excellent explanation! thanks for your generosity!

[@markoshun]

Nice! I think I could even explain the gist of it now after this demonstration, which means I actually understand.. finally. Thanks much.

[@StevenMorris]

Excellent video! Thank you for this clear explanation Gaz :)

[@highpeakrecordings]

so useful Gaz! cheers.

[@mrmorpheus9707]

You mean mix In 32 bit float. or just export. at. 32bit float?

[@codytj3434]

export

[@greedokenobi3855]

He will reply to your comment soon. Give him 2 more years lol.

[@christoroppolo8742]

Gaz. I got so fed up with Studio one Pro because the updates that did nothing for my work flow and the “pay for” library stuff . So I am now using a combination of Gadget for my iPad pros and my iMac pro quad core 2015. Studio one was so buggy in the beginning that it drove me away from the proper computer daw and straight to iOS which I fell in love with for many reasons . But what do you think ? Isn’t it time for the Apple UBER IPAD PRO? The actual equivalent of their flagship desktops? And the insidiously annoyingly clunky laptops? (I hate the flip screens etc)Peace Christo 👽🔦✌🏻🙏🎶👏🏻🛸🕺👍🏻❤️

[@giulioboobzilla]

Christo Roppolo What does Studio One has 2 do wit da subject of dis video?

[@piscue]

Do you use this 32bit also for releasing stuff? Do you worry about bit reducing when you convert to 24bit/16bit to match the format you will? Any thoughts about dithering?

[@ronrobins3513]

Great idea can't wait to try it

[@FarrellMcGovern]

Nice tip!!! My only question is...does it work across all platforms and software? Or is it just a nice feature of Cubase?

[@GazWilliams]

Other software lets you create 32 bit float mix downs but some won't show you the peaks once you have reduced the gain, it may still appear to be clipped but isn't. I think that is down to some software having dynamic graphical updates to the wave rendering and others being a bit lazy and not re rendering once gain is reduced!

[@FarrellMcGovern]

+Gaz Williams Great! I will look to see if my DAW, Reaper, will allow me to do this! Thanx!

[@giulioboobzilla]

Farrell McGovern I hope Reaper does. Cubase since SX1. Logic doesn't support it. New ProTools does. Fruity Loops, 2.

[@DoctorJezz]

Gaz - what VU meter levels would you suggest to mix to pre the master bus and in the aux sends (it's that digital gain staging question again!)? Could be an entire presentation for you, given the length of some of the forum discussions! Great work anyhow :)

[@Robstafarian]

Thanks, Gaz!

[@Pharesm]

This is great, but what kind of mixer can I buy that supports that??? I only know a few very expensive portable field recorders that do that...

[@annother3350]

You can downsample the file once you're sure it's ok. Use Adobe Audition, Sound Forge or something

[@ikamy]

Hi I have a project 48k 24bit, when vocal files arrived they are 44k 32bit , should i upgrade the project to 32bit or should i convert vocals to 24bit, and regardless of the solutions at some points i need to convert 32bit to 24bit How can i do that with minimum loss? Thanks

[@giulioboobzilla]

Cameron Bright Da problem is 2 match the 48/44.1kHz. 24/32bit is no Problem. Cubase can handle both in da same project

[@dallasphill]

Convert the 44.1kHz vocal to 48kHz

[@j7ndominica051]

It would be cool if loudness war casualties could be restored in level with peaks intact from digital files (mp3, aac) that have internal range of over 150 db. Apart from WAV, Reaper supports rendering over full scale to WavPack and MP3. Other formats currently clip. WavPack files are smaller and be selected to have noise floor of 16 or 24 bit equivalent to reduce their size further, while still encoding overs.

[@reignz1397]

This can only be done through mixdown or can a you start a new project in 32 bit float?

[@marktywharton]

interesting - it's the floating point that makes the difference? what happens to the 32 bit float mix if you then convert to 16 bit without changing the levels?

[@GazWilliams]

The 16 bit version would be clipped unless you reduced the gain whilst still in 32 bit float. Also dithering is necessary with bit depth reduction as I'm sure you know.

[@ODWALLA123]

Thanks, Gaz.

[@dcashley303]

Wow, information like this for someone without a professional background in production is like gold, thank you Gaz. Keep them coming. A quick question, does that also mean that if you record at 32-bit you are less likely to get clipping? (I imagine not but wanted to be clear).

[@janminor1172]

I don't think you can record in 32bit as no audio interface can deliver that kind of dynamics. I advise recording in 24 bit and keeping an eye on the level, don't record too hot, keep it at around -18dB RMS and peaks at maybe -6dB peak, and you are safe. Also, if you have those levels before you go into the mixer and plugins, you reduce the likelihood of clipping the output and most analog emulation plugins today are programmed to be used with a -18dB RMS reference level anyway.

[@dcashley303]

+Jan Minor Brilliant, thanks.

[@giulioboobzilla]

Jan Minor I always recorded in 32bit float since Cubase SX1 and itz amazing how much headroom U got

[@takeshi_shinyama]

Appreciate it sir😌🙏

[@annother3350]

psst -- a handy little tip -- I've noticed saving mp3s from 32-bit files to 320k mp3 keeps the full high end!! Using any lower bit rate results in very top end completely chopped off

[@GazWilliams]

Ann Other aha I’m going to have to try this out. If this is so, I would wonder how it affects other codecs.

[@AkkiMusicGh]

I just subscribed I think I have found my real teacher finally

[@HugoDaSilva7]

really good explanation...thanks

[@annother3350]

it's also better for avoiding artefacts when using master buss plugins

[@johnnypunish]

Explained well, thx! Well Done!

[@oinkooink]

But what if you mixed down with a true peak limiter making sure nothing goes over -1db?

[@okaravan]

-1 dBTP seems to be an attempt to solve the problem in a place where it doesn't exist. When audiofile is normalized on 0 dB in a normal way, it is not distorted, it is perfect. The fact that analog signal can go higher than 0 dBFS during playback is not a problem, it is part of the digital-to-analog conversion process. And output limiter is also not essential, if it is not achieving some artistical goal or solving some technical problem or simplifying workflow. 32-bit float export allows not to worry about clipping and quantisation noise floor. But after this export is done, you can open this file in an audio editor, normalize it on 0 dB or -1 dBTP, and then save as 24-bit FLAC. After that 32-bit file can be deleted. When output limiter is used, it is enough to export directly to 24-bit FLAC.

[@100chuckjones]

The logical question is this.. What streaming sites except 32 bit float point files ? Umm.. None. So what are we gaining from exporting to 32 bit float point when you are eventually going to have to dither it back down to 24 bit eventually ? I run a facebook group with over 30,000 users who keep referencing this video as a reason to mixdown to 32 bit FP but don't understand what they are achieving. Appreciate your efforts but as a mastering engineer, there is no real value in exporting to 32 bit FP unless you are exchanging files with other musicians (the stems).

[@GazWilliams]

Good (floating) point! My main reason for this was more for never ever clipping your mix providing the mastering engineer knows to normalise to -0.3bB or so.

[@vellannoorbalan6353]

Sir, so what can we do to write CD. is it support 32 bit?

[@giulioboobzilla]

Vellannoor Balan Not an audio CD. itz limited 2 16bit/44kHz

[@380stroker]

Good say sir. Jolly good day.

[@ronrobins3513]

question ...is this also available in Logic Pro X

[@giulioboobzilla]

Ron Edwards Unfortunately not :( Logic doesn't even support mono export 4 single file Bounce. Cubase has 32 bit float since SX1 I think. Fruity Loops supports it and da newest Pro Tools.

[@JMEDIAUK]

Thanks this video really help my end product out really well u gained a sub 👍

[@FoundationStepper]

So helpful ! 🙏

[@Alexander-mk4qf]

32 float = huge volume workspace 16 = classic volume workspace

[@davefk]

Interesting how those extra 8 bits help out, but I wonder what the waveform would be like if you normalized it?.....

[@s1gnal_]

Great job! Although your mic level is waayyy too low compared to the mix.

[@GazWilliams]

Sorry! Thanks for pointing it out, I am using new software for this series so I am learning as I go!

[@s1gnal_]

+Gaz Williams No problem! Keep up the good work! :)

[@corystone5248]

how does this translate on cellphones and mobile listening devices?

[@380stroker]

Like a potatoe

[@steveg219]

this is great info, thanks!!!

[@YOURGEN-q3z]

thanks a lot i must work on a cd who is burned !

[@SamsonMusicStudio]

awesome man!

[@ishaandobhal6397]

you answered my question. Thanks.

[@elpablitorodriguezharrera]

Is your session 24bit then mixdown it anyway at 32bit float or it’s 33bit float from your session?

[@380stroker]

No, it's potatoe.

[@rohitRraina]

Woah! Awesome ❤️❤️❤️

[@youngworld001]

I nd held I am recoding in adobe addition 3 with my 2i2 scarlett interface witch is 24 bits but my software only hv 16 and 32 bit witch should I recording ?

[@edmoss2020]

Thanks for this video....very helpful.👍

[@hakuu2559]

What if you drop all tracks in your mix down so that your master is peaking at -6db. Would there be a reason to export at 32 bit at that point?

[@PureAmbientDrone]

Hakuu my thoughts exactly.

[@triniticoclough6387]

Yea I would mix and have my gain stage set so that my mix sits at -6 or at the very most -3 giving mastering head room to do his or her work. Then mix down /export ( faster than real time bounce/ render) at 32bit 48k just to ensure there are no truncations or clipping of the audio file. Any Mastering facility should be able to handle a 32 bit file..if not..I would also provide along side the 32 bit file...the same file converted to 24 bit 48k in hopes that the the transition from the 32bit file to the 24 bit file would leave everything in tact. I would prefer for them to use the 32 bit file and master from that insuring they are getting the exact sum of my mix to the point if they was listening to the file vs the session … there would be no difference in terms of the mix itself! However in the past I have delivered files 24Bit 48k coming from sessions mixed at 32bit 48k while keeping my gain stage set at -6 / -3. No one has ever complained! I know many folks are told to keep their master buss at 0. I have never done this. Maybe I am wrong but the checks keep coming so hehe..

[@greedokenobi3855]

Channels that respond to their viewers questions are the channels I’ll never sub to. It’s been a year and still no reaction. That shows you how much they really care.

[@annother3350]

Hakuu -- the advantages of 32-bit are many. The headroom is one. The fact you'll get less artefacts from mastering plugins and also saving your 32-bit file as 320k mp3 wont chop off the high end at all!!

[@380stroker]

No.

[@nathanbanda8218]

Amazing

[@NicoGotTheSauce-]

🙏🙏Thanks for this one🙏🙏

[@canvoodoo]

Great vid thanks

[@RadiAsian]

thank you

[@moreausylvain]

0:47 ... wait what ? you don't check if your mix is kind of clipped ? like... these red lights set to infinite hold.. how do you miss those ?.. why am i even this video ? :'D

[@annother3350]

many people after working on the track for a day may make a last minute tweak and not bother to play the complete file again -- especially if it's an hour long

[@macrondo5852]

so 32 bit is for people who can't gain stage correctly?

[@giulioboobzilla]

mac rondo Not really? How can U gain stage perfectly when U don't know how da singer will sing during recording. If he gets 2 loud U will hav a problem wit 16/24bit, not wit 32bit float tho it has also limits, maybe 10/12/15 dBs. Check it out!

[@380stroker]

Yes. People who can't gain stage and are rookies will use 32 bit floating point mixdown. So many people in the comments don't even know the basics such as integer vs floating point. They are not aware that their daw mix engine is by default 32 or 64 bit floating point math. Many don't know what a mix engine is.

[@ComputingSound]

Top tip!

[@soundsfromYYBY]

why record hot anyway no matter what you recording in lol

[@Hitsmithandmusic]

THANK YOU SIR!!!

[@timbalandhits8]

can I record 32 bit float in logic x??? any body help?

[@giulioboobzilla]

don candiani Not yet. Cubase since SX1 Fruity Loops. New ProTools

[@FSmith-gz1dh]

thank's Papa Smurf :)

[@FSmith-gz1dh]

now I´m more confuse

[@rollacoastaride1937]

cool, cheers

[@khoimckinney1302]

You are the best

[@ivarisaksen]

Thank You :-)

[@babblfish8038]

But you need an interface that can handle 32 bit ...the most have just 24 and that is enough

[@380stroker]

You are confusing integer and floating point and DAW mix engine resolution.

[@EfisioCross]

❤️

[@pcuimac]

thx for the good tip! As others already said, your mic lvl was to low. ;)

[@GazWilliams]

Teething trouble but thanks for the advice!

[@dumisanitsotetsi2272]

Forgot to like and subscribe

[@jurajzenhanakovic]

Oh the irony = expert on audio and his video has set audio volume extremely low (his voice, not music). Video side - very overexposed areas (face, keyboard) aswell. Good info tho.

the audio is very clean and balanced, not too low. Your ears must be the problem. His video ? Hotspots? I don't think he cares..this isn't about video. I'll go judge your videos now. Oh of course 'NO CONTENT" Joker. What happened to your dome? You lost some hairs.

[@annother3350]

@ he could have turned the distorted bits down -- woke next doors cats up

[@foundsoundsmusic]

Ironically, the audio in this video sucks.

[@rsvp9816]

cant listen to someone who doesnt normalize their video levels.

[@annother3350]

too late

[@ComputingSound]

Top tip!