I am reducing the bit depth of some FLACs from 24 to 16bit (maybe keeping 24, I am not sure) and the sample rate from 192k to 48k (per research it’s suggested to reduced in common divisors, so div by 4 is 48k).
The situation is that I am not sure what kind of dither to apply as per research it shows that the less intrusive distortion should be the triangular with shape option, however I have no idea about what shape form, extended shape form, intensity or if I should use adaptive mode.
I hope any expert in the community can give me some suggestions or explain a little more about these options to keep these FLACs as good as possible. Thanks!
What software are you using?
My original idea was to try checking if I can notice the difference between one of those massive 24-192k or even f32-384k, and some more conventional 16-44.1k or 24-48k. But after some checking around I got stuck in the dither option and I found a little hard to understand the algorithm differences.
So basically it was a small experiment and since apparently files over 24-48k are suppose to be non distinguishable for most humans (if not all, but that is probably a can of worms I don’t want to open), I found those massive hires files to be a waste of space for no particular gain other than brag about how many TBs of hires music I have. Anyway, I am still learning :D
Why man? Just get mp3 or wav.
There is literally zero reason to use WAV over FLAC, it has no compression or native tagging support.
Well I’m not so informed about FLAC, but he wants to save space so cd 16bit 44.1khz quality in what ever file format I would say. But better get some storage
If you keep the bit depth at 24, I wouldn’t bother about the dither. It is at -144 dBFS so no playback chain is able to make this audible.
If it is 16, dither is considered a must. Can’t advise you on the specifics.
Thanks for the great insight, it is interesting to learn about these details. I think that of I keep the FLACs 24 bit-48k it might not be the dither as you mention. Maybe dither becomes a lot more useful on some situations like podcast production to avoid the extra noise while keeping a relatively small file size.
Shaped dither is meant to move more of the energy of the noise into the upper frequency bands, and thus be less potentially audible to most folks. Triangular dither has a noise distribution (so how the random values of the dither are spread around the true sample value) that is more often close to zero than a uniform spread, though I don’t know why it would be superior to Gaussian.
Second the vote to just keep it at 24 bit unless space is a real concern. If it is, shaped 16-bit should do fine.
Bruh… Dithering not needed for a 24 but file… Who’s gonna hear -144dB?
Monty from Xiph is on record saying that even no dithering is perfectly fine for 16-bit audio. This is probably true. The quantization noise is really quiet, you simply aren’t going to turn your equipment up enough to hear noise at some low level like -96 dB, or wherever the quantization exactly ends up being. There is probably enough background noise in the signal to begin with, to make the whole question utterly irrelevant – if there is already a broadband hiss at, say, -80 dB, it hardly matters if you add tiny amount of additional quantization noise at some -96 dB level. It’s just going to vanish there and you can never tell it was present in the first place.
That being said, I do recommend use of dither noise, and triangular dither is not bad. These settings you have there control the noise shaping, which can somewhat extend the retention of signal at hearing-criticial 1-2 kHz region at expense of more noise in less-critical bands. I think any choice, including “None” for flat dither noise, is perfectly fine. U shape dither is likely technically the best choice, suggesting by name that it places dither noise at low and high frequencies roughly following ear’s sensitivity curves. If you had some signal near, -96 dB to preserve, then you would likely want to make this choice to ensure that as much of the audio in the hearing-critical bands would be preserved as possible.