AudioSuperResolution

SpectraLayers
1

How about adding a AudioSuperResolution ML feature to upscale 44.1 KHz 16 bit recordings to a higher level by enhancing the missing frequencies

Like Audio Super Resolution

3 Likes
2

That would be awesome!
Somewhat related to this subject.

2 Likes
3

Wow! I need this

how do we know if the source is aliased or LPF applied? My recordings are DAT 32KHz that I’d like to upscale…I haven’t done anything to the recordings other than transfer to computer

4

There’s another project that concentrates on video conferencing:
NVIDIA Maxine
The SDK includes an upscaling technology but it is limited by the audio resolution and as far as I already tested it’s restricted to mono audio

But it’ts easy to create two mono files of a stereo file :wink:

1 Like
5

Yeah! I agree but again, the foundation for achieving this is kind of backwards. It’s like teaching a new pilot that in order to fly an airplane you must first learn to use a submarine. In other words, it doesn’t make any sense upscaling Wave Files!

1 Like
6

I agree…unless we can AI to get to some form of re-amping

is it really impossible to realize a better freq response than say a 32KHz/11bit audio recording? I don’t know…yet I never thought I’d see the day when peoples’ voices can be cloned and then fed text…but here we are…and it works very well IME (we’ve been using it a lot in the past 12 months)

7

Just for clarity sake

Are you are asking for two independent but seperate solutions

Sample Rate Conversion and Recovering or resynthesising lost or missing high frequencies (resolution)

Having them both in the standalone would be great.

As a side note: SRC is done on the fly by Nuendo with extremely good results.

Other DAW too have some form of SRC implemented.

1 Like
8

I’ve always heard “let the hardware do the SRC”
but maybe that is an outdated solution when considering current tech? Further, some of those tapes are too degraded to try another transfer.

In my DAT recordings, I can tell the difference between 48 or 44.1 or 32 KHz

32KHz being obvious that freq response is limited…and a great deal of the material is 32KHz because it was the LONG PLAY setting

I wouldn’t say specifically that I can tell it sounds like a LPF…looking at spectrograph scales, though, it is the high freqs that are less than 44.1 and 48KHz recordings :slight_smile: