Hi Urs,Urs wrote: Thu Oct 19, 2023 6:40 am Let me preempt high expectations here.
There will not be any dedicated resynthesis or sample playback. The tools we are building are optimised for human interaction, which I think offers spectacular results. Any form of sample import is a matter of convenience, but resynthesis or sample playback was never the aim of these tools.
While we did recently show that we can analyse samples and generate vectorised curves from them, I think I was also clear that it's a bit hit and miss. Some material is a better fit for this than some other. Importing individual waveforms is pretty easy and straight forward. Importing and vectorising a dozen or so frames from even a simple sample does not necessarily produce the expected results when morphing: The transitions may be seamless, but they hardly ever sound exactly like the in-between frames of the sample. It takes some manual work to achieve this, and often enough it simply isn't possible. It's pretty good when it happens though, and then I guess it's safe to say "this is a bit like resynthesis".
However, converting arbitrary samples to spline based waveforms resembles wavetable synthesis, not resynthesis.
We have also planned to support classic wavetable scanning like Hive does. We furthermore have plans for Hive and Z3 to import samples of arbitrary pitch, which internally will be converted to wavetables with up to 256 frames. Those will have the same properties as any wavetable, it's just an easier way to create a wavetable from an arbitrary sample. It is my understanding from what people tell me that the tools available for this are not exactly gratifying, which is, I guess, why people frequently ask for it.
This again is wavetable synthesis, not resynthesis.
What we will offer in Zebra's new oscillators though is pretty extensive manipulation of harmonic amplitude/envelope and pitch/frequency. But this again is not accessible on a per-harmonic basis. Instead, high-level oscillator effects and spectrum warping offer realtime manipulation through modulators. These tools can do some crazy stuff.
In any case, we will not be advertising any of it as resynthesis or sample playback.
I've been following all of these threads and have been waiting for the Hive 2 improvement that takes a sound byte and turns it into a wave table. I see that you've mentioned it here. I'm not clear from what is said above how things are likely to work with this. Part of what you say sounds like you've had difficulty with the results of doing this--hit or miss. In another part in the post above you mention importing "samples of arbitrary pitch, which internally will be converted to wavetables with up to 256 frames". So it sounds like a sample based wavetable creator tool will work to some extent.
Maybe the part I'm not fully understanding is the Wavetable vs resynthesis differences mentioned above. To me, if a arbitrary sample of say, a 5 second morphing bell sound to wind sound is converted to a 256 frame wavetable, and the sound from the wavetable sounds the same as the sound from the sample, then essentially, that's what a person would expect from resynthesis. Are you just saying that even though the converted results end up as a wavetable, it is still not resynthesis, and this is why you are not advertising it as resynthesis? To me, I'm less concerned about what it is called, than the results in the end. I'm just trying to understand how accurate the tool will be with converting the sampled sound to a wavetable that sounds the same. It sounds to me like this process is still hit or miss. Could you clarify and explain further? I'm seeking to understand. Thank you!
