As you all know, all files has to be in 44.1 kHz when uploading music.
I was just wondering if you feel there is any audio quality loss if you’re mixing in 48 kHz and 24-bit and then downsampling to match the file requirements? I’ve looked around trying to find a good answer, but couldn’t really get one. To be honest, I don’t really hear much of a difference myself, but was curious about what you guys think?
I know that upsampling is something else, because you are trying to achieve a higher quality of file which can possible affect the audio itself, but I am not so sure about the other way around and not experienced enough to really know about it!
What you can do to preserve some quality is put a dithering plugin on the master bus. When exporting from a 24 bit into a 16 bit there is some loss and this could possibly result in some artefacts. What dithering does is put a very low level noise into the track and by doing that it masks the possible artifacts.
This is a very compact explanation but you can find more about it online.
I am mixing in 48-24, than when final mix ready i am switching project to 96-24 and render 44-16. A little bit better overall clarity in final version.
With huge amount of quality VST instrument and mastering and low latency, mixing on 96-24 is too demanding for CPU (though I have Ryzen 7950x). As I did some subjective check, 96-24 to final 44-16 sound less harsh on highs than 48-24 to 44-16.
I find it true that some reverbs and effects sound and process better in 96khz even after exporting it back to 44.1
I read an article about it somewhere but forgot where.
I recommend repeating tests, this time with objective blind test, preferably involving other people. My knowledge and experience tells me that they should not hear the difference (let’s say I am 99% sure). You can even post A/B files here.
Besides that why do you mix in 48 besides 44,1? Why? These few kilos doesn’t make differenece at all. The difference between 24 and 16 bits too.
it seems you know better than me what good and what not. As for me - as @CWMusic already explained - reverb and delays sound cleaner, less harsh, 48 i use in my project as default because i am recording electric and now acoustic guitars too , as well as some hardware synth. All in all i can hear the diffrence in my workflow. I have acoustically treated studio with great sound, custom coaxial monitors with great detailed sound and in my setup the difference is obvious. No time to argue.
If we are talking about specific sound design tools like pitch shifting of samples recorded with microphones like Sunken or Sennheiser mkh 8000 - totally agree. However this is a relatively rare exception for composers (though common among sound designers.)
But if we are talking about general better quality, e.g. of reverbs: let’s make blind tests, convince me. Without this, it is only an audiophile urban legend for me
Sorry, didn’t want to sound harsh. It is a topic which fascinates me and I spent decent amount of time experimenting on it.
I’ve made enough blind tests which don’t allow me to agree that higher resolution is better “in general”. Besides sound design of specific samples mentioned above (that’s why I sometimes work in higher resolutions).
Even though imho there could be audible difference between 88,2 or higher and 44,1 (in the mentioned sound design exceptions). But not betweeen 44,1 and 48.
Prove me wrong. I would be over the moon if I could finally hear the difference in the blind test.
PS: I would love to hear a blind test of mentioned reverbs. In theory specific reverbs could sound differently (maybe?). But still I doubt if the potential difference is either audible or positive.
Why don’t you work on 44,1 and then move to 88,2 and back to 44,1? Multiplying and dividing sample resolution by two is the only natural convertion.
Converting from 96 to 44,1 by fraction introduces kind of “artificial” tones in theory.
In other words I would chose these workflows:
44,1>88,2>44,1
or
48>96>48
PS: I doubt if this converting has audible reason at all but I always thought that this is how it should be done to avoid theoretical convertion errors.
Nobody uses such a wide dynamic range which could generate audible “tail” artifacts you are mentioning.
These few kilos doesn’t make differenece at all. The difference between 24 and 16 bits too.