Could something similar be implemented to SL?
this does very much interest me as I have a lot of 32KHz Sony DAT recordings of live improv music and listening back, that hi freq loss is quite noticeable. Back in the day (the 90s), DAT was expensive and we often used “Long Play” not realizing the sample and but rates were so drastically reduced…I think Sony’s LP was 11bit, that I just save as 16. Would the upsampling improve the bit rate too?
Further, the great thing about using a a tool like SL is you can check your hearing. I found I can’t “hear” any freqs about 11K; my wife claims she can hear up to 13K. I say that because the video freq scale showed no info above 15K, but in the deffierence signal of the upconverted audio file, we saw improvements from 9K, which I, personally, would notice.
That said, I currently have a 32KHz recording loaded in SL11 and the spectrogram shows info up to and beyond 15KHz…so I remain confused 
That is far too involved for me but upscaling is indeed an interesting subject!
Do you have any opinion about the PGGB way of upscaling?
I mean, obviously you get a higher bit resolution - which perhaps could be a good thing when applying e.g. Unmix Song for better stem separation in SpectraLayers - and also a higher samplerate if you want to.
For listening purposes that would yield a need for special DA converters I suppose, but maybe one could use one’s regular 24/96 converters and apply the SpectraLayers’ Unmix magic without playing the file through the DA converters, that is, ’in a silent way’, and then downscale after the unmixing? And hope for a better stem separation result…
This is a totally new domain for me so I might be a bit beside the logics here…
1 Like
yeah, that vid is not my domain neither!
The more you have information the better the results on unmixing. You can see it, by comparing mp3 and wav file of same audio. Some historic recordings have been crappily transfered to digital domain, so this could help in situations like that
1 Like
ah, I see…makes sense…I never work with mp3/ogg other than downconverting for online
So…I just opened a file from my documentary I’m working on.
Freq scale shows info up to 25KHz…these files were recorded at and I use them in DAW and SL @ 48/24.
The music I was talking about earlier was recorded at 32KHz/11bit and I added zeros to the headroom by transferring from DAT and saving as 16bit. That music shows freq info up to 16KHz.
I know I can’t hear much if at all above 11K; BUT, I can hear a difference between 44.1K or 48K and 32K files. I seldom use any higher sampling rates. Most of my DAT recordings are 48 or 32K…I don’t work with 44.1KHz very much other than to render to or listen to commercial music.
Still learning to understand the spectrogram
Yeah, I’m aware of that, and Robin Lobel has told unmixing a 24-bit song would give better results than unmixing a 16-bit.
Hence my speculation that unmixing a song prepared in PGGB would give even better results. PGGB can upscale to fantazilious 64- or 256-bit depth…
1 Like
Actually!
Audio (in-general) is kind of an outdated standard. Meaning an image file(a spectrogram/spectrograph visual representation of an audio file) is superior to an audio file in a lot of ways. In terms of upscaling, there are more algorithms(research and development) that have gone into upscaling images and video than in audio… While kind of subjective, one can argue and agree that there is a lot more information in a still image than an audio file (depending on file sizes and whatnot).
The problem with audio is that there has never been an official image format for audio. Meaning there is no standardization of image files for audio. 99% of the recording engineers who record vocals are most likely to record vocals as a wav file as-oppossed-to a image file (because there is no format for recording audio straight directly to an image file). The format is old and to upscale an old format like that wouldn’t make sense (because you’re working with old philosophies and old outdated ways of thinking). Something like scaling can be done but it would only make more sense to do it with an image file as-oppossed-to a wave file.
2 Likes