Sample compression
-
I just came across this - http://www.monkeysaudio.com/developers.html - would it be a practical solution for implementing sample compression in HISE?
-
Yes I heard of it but its slower than FLAC (according to some benchmarks I've read) and FLAC is already too slow to stream huge amounts of voices. But I might be wrong about this. Have you any resources with other benchmarks?
-
Actually I think you're right. The only info I'd looked at when I posted this was the Monkeys Audio site, I've since looked at the wiki and some other general info sites and it definitely seems like everyone thinks it's slower than flac. What do NI do to get such good compression and high speed I wonder...
-
Pretty sure Kontakt uncompresses to a temp file as it loads banks, thats why loading is not particularly fast on Kontakt, I could well be wrong on this though.
-
Ah that's interesting
-
Are you sure about this? This sounds like cheating :)
-
What sort of CPU hit will I get using FLAC instead of WAV?
-
Factor 10. Makes no fun. It also seams to use a global state so decompressing multiple files simultaneous brings an extra hit.
But I am thinking of writing a custom compression format in the not so distant future. Monophonic instrument samples do have lots of redundancy and I think an algorithm that takes this special type of material into account can beat a general purpose codec like FLAC...
-
That would be great. I've built a small patch with FLAC samples and didn't notice a performance hit but it didn't have any scripting and only one articulation, I wouldn't like to risk it on a larger instrument.
-
You can simply convert all your samples to FLAC, copy the HISE patch into Sublime, replace
.wavwith.flacand reimport that into HISE to compare the performance. -
That's clever! I'll give it a go
-
Did you maybe try this out with a larger instrument? Interested to see how it compares in performance
-
@Dorian : I wrote a custom audio codec that outperforms FLAC by a factor of 10 - see: http://178.62.82.76:4567/topic/236/hise-lossless-audio-codec-is-ready
It doesn't compress as good as FLAC (you can expect the file size to be reduced to 45% instead of 25%), but it has hardly any overhead compared to uncompressed files.
FLAC in it's current form is not suitable for streaming multiple voices (it allows ~70 voices to be decoded in realtime on my system vs. 700 uncompressed / HLAC compressed voices)
-
this looks great
