Official Serum 2 thread!

[tumface]

Wave shaping is any mathematical process in which a wave is processed into a distinctly different shape. It is achieved by applying non-linear transfer function to the original waveform. That function can be simple, or very complex. Folding, wrapping, and clipping are all examples of wave shaping. It's not just mapping one wave onto another.

I know what waveshaping is. You can't create a multi-voice effect from an arbitrary waveform using a waveshaping function, right? The waveshaping function is stateless.

[teilo]

Wave shaping is any mathematical process in which a wave is processed into a distinctly different shape. It is achieved by applying non-linear transfer function to the original waveform. That function can be simple, or very complex. Folding, wrapping, and clipping are all examples of wave shaping. It's not just mapping one wave onto another.

I know what waveshaping is. You can't create a multi-voice effect from an arbitrary waveform using a waveshaping function, right? The waveshaping function is stateless.

Correct, using a single function. It would have to be combined with time domain techniques.

[tumface]

Correct, using a single function. It would have to be combined with time domain techniques.

Which ones?

[jrwaltb]

I am trying to see if I am misinformed or whether there are any official updates and promises that I am not aware of.

I know the AMD chipsets are newish but its still very different low level code that creates the audio features Serum2 needs to use so its best just to ask the guy running the project. Also, saying resampling and quality options are nothing new you are assuming Serum2 is worse than phaseplant/etc in terms of CPU witch it is NOT for many people and many synths do NOT contain the resampling options as Serum 2. (Hive, PP, Icarus3, Melda) Just talk to Steve about it if you already own it.

[Echoes in the Attic]

Finally got around to demoing this. Wow this thing really has everything. Except one thing. MPE doesn't work properly playing mono legato. Play a note, glide up, kill that note by playing a new note and the second note will be wrong. It will be pitch bent according to the previous note glide. Common problem, but worth fixing.

[swilow11]

Wave shaping is any mathematical process in which a wave is processed into a distinctly different shape. It is achieved by applying non-linear transfer function to the original waveform. That function can be simple, or very complex. Folding, wrapping, and clipping are all examples of wave shaping. It's not just mapping one wave onto another.

I know what waveshaping is. You can't create a multi-voice effect from an arbitrary waveform using a waveshaping function, right? The waveshaping function is stateless.

Reasons Malstrom synth has some wavetables (graintables) such as the saw16 one where increasing the scan speed genuinely sounds like (quite bad) unison. You could also use fm from a chaotic noise source at very low levels to create a sort of pseudo supersaw sound.

I still don't think there is any evidence that Hive does this. If you alter the phase reset options in Hive you get precisely the behaviour you'd expect- with Phase retriggering on, you get the typical transient bump as all voices start at the same time. If you randomise phase, you get the typical dispersed inteference pattern supersaw sound. I guess this behaviour could be modelled but that actually seems more complicated than just duplicating voices as one would expect.

Although Hive actually does have a pretty unique sounding unison but using Bitwigs stack spread modulator does actually seem to reveal multiple seperate voice paths that you can send on their own little modulation journey.

[nusound mind]

That's a silly take for two reasons:
1. There are many synths with multiple quality modes like real-time, rendering etc... It is nothing new in the VST world
2. Ryzen CPUs from the last 2 generations are not old chip-sets.

I am trying to see if I am misinformed or whether there are any official updates and promises that I am not aware of.

One of my cpu's has one of those Ryzen chips and Serum 2 works fine on it. Yes, there are some cpu melting patches but they're few enough, and that can be said of a number of softsynths, esp. newer ones. Ofc. I can only speak for my own setup.

[SirKen]

I know the AMD chipsets are newish but its still very different low level code that creates the audio features Serum2 needs to use so its best just to ask the guy running the project. Also, saying resampling and quality options are nothing new you are assuming Serum2 is worse than phaseplant/etc in terms of CPU witch it is NOT for many people and many synths do NOT contain the resampling options as Serum 2. (Hive, PP, Icarus3, Melda) Just talk to Steve about it if you already own it.

What?! I am trying to get info before the intro price period runs out. I am trying to make a decision on the current state of things and/or trying to figure out whether there is any hope for future improvements. I am moving out of self interest.

Now you tell me why you responded the way you responded?

[SirKen]

That's a silly take for two reasons:
1. There are many synths with multiple quality modes like real-time, rendering etc... It is nothing new in the VST world
2. Ryzen CPUs from the last 2 generations are not old chip-sets.

I am trying to see if I am misinformed or whether there are any official updates and promises that I am not aware of.

One of my cpu's has one of those Ryzen chips and Serum 2 works fine on it. Yes, there are some cpu melting patches but they're few enough, and that can be said of a number of softsynths, esp. newer ones. Ofc. I can only speak for my own setup.

Thank you for the input. May I ask what chip do you have? I am on 5900X (Zen 3) and it was somewhat usable with a premium on the CPU for most patches except a few offenders like you said. I think Granular oscillator might possibly be the worst one. Were you able to narrow down the offending features?

For a similar reference, I had issues with UVI's Falcon, specifically with its stereo spread and unison features, while I was running my DAW in 96K. The sound was crackling even when the CPU meter was at 40 to 50%. It all went away when I switched to 48K. I am wondering if there is a specific knob or a button that I need to avoid in Serum 2 while the performance hopefully gets improved.

[nusound mind]

I didn't get too in depth with it but from what I recall, too many oscillators, voices, too long of a release, and yeah the granular is a culprit as it seems to be on other synths. Unison also a factor.

I do hope it's optimized further to some degree in the cpu dept. but it's still quite usable for me as is, just not in all it's maxed out glory.

[jrwaltb]

Now you tell me why you responded the way you responded?

I am in this conversation because I am trying to explain that the resource issue does not exist for me and for many other people. I have no idea why, but I know Steve has personally talked about it, and I am sure it is not being overlooked, as it must be a specific chipset issue. If you have a chance to get it cheaply, I would 100% recommend it. Every time I use it, I create something really good and I am not even using the multi-sampler or spectrum features yet, as they are overwhelming. People who actually know professional sound design must be making some incredible things with this synth, which I am sure they bounce to audio and then rinse and repeat.

[kraster]

Personally my experience with Serum 2 and CPU issues has been that it has been taxing on the CPU when you'd expect it to be.

A 7 part unison Granular oscillator being FM'd by a modulated wavetable is going to tax any system.

On my system Serum 2 is slightly more demanding on the CPU than equivalent synths but not markedly so.

It's not like it's buckling when playing wavetable or sample based things.

[SirKen]

I am in this conversation because I am trying to explain that the resource issue does not exist for me and for many other people. I have no idea why, but I know Steve has personally talked about it, and I am sure it is not being overlooked, as it must be a specific chipset issue. If you have a chance to get it cheaply, I would 100% recommend it. Every time I use it, I create something really good and I am not even using the multi-sampler or spectrum features yet, as they are overwhelming. People who actually know professional sound design must be making some incredible things with this synth, which I am sure they bounce to audio and then rinse and repeat.

Are you on Ryzen?

[Ah_Dziz]

Correct, using a single function. It would have to be combined with time domain techniques.

Which ones?

Modulating the shape of the transfer function would do it (in the most simple case you'd be modulating the "phase" of the transfer function). Biasing the phase or zero crossing of the input to the waveshaper (waveshaping here meaning: a transfer function that maps each input sample level to an output sample) with an lfo or phasor. Pre and post filtering ( single pole) with cutoff modulation of the input of the waveshaper or or of the transfer function ( the latter would really just be averaging the array data in some dynamic way).

A rising saw (ramp) waveshaper function gives you the same output as the input. Anything you do dynamically to that saw will give you a very different output that changes over time. So by taking one oscillator and multiplying it through N different versions of an array you end up with N different versions of your oscillator. You can get tons of neat effects with this technique.

[tumface]

Sounds reasonable. I've never tried that for unison. Though, I dunno if I'd still refer to that as plain 'waveshaping' in this context. Usually I see it called transfer function, and if it's exposed to the user of the synthesizer for sound design purposes, then it's called waveshaping. Maybe I have a faulty memory, though.

(I think I remember working on something for a unison wavetable oscillator in the past and, IIRC, I precalculated some stuff that I used to offset the tap locations for sampling from the wavetable, before filtering and combining the results. I don't remember exactly how it worked... I think I read it from a book?? Does that sound right?? I might be tripping.)