Distressor
-
Hi, can anyone help with creating an Empirical Labs Distressor-type in ScriptNode? I'm happy to share the result with the HISE community, which would be a multiband implementation.
The best Distressor plugin from Slate Digital is the best (with apologies to UA). If anyone helping would like a subscription to all the Slate Digital stuff to get a feel for it, I'll set you up.
-
Thats a coincidence: I was planning on doing a distressor properly in the near future, but I didn't have a unit on loan. I would've wanted to definitely take measurements from the real thing. I've already done preliminary studies on the circuit though, it's very similar to a 1176.
If your goal is to approximate it from existing plugins, that sounds interesting. I've been playing around with compressor algorithms on and off over the last few years.
What did you actually want help with?
My approach wouldn't be to use purely existing scriptnodes, but to use c++ custom nodes. All the best models of this compressor use circuit modelling in some form, I'm not sure how close you can really get without it. -
@griffinboy I was thinking Distressor-like. If you want to do it exactly, I can get you a loaner and some support…but I'd ask that you get a license from them (for the algorithm, not to call it a "Distressor"). I can help with that.
-
You've definitely got my interest...
Can you help me understand what you are proposing here? I don't entirely understand the post!The gist is that you are looking for someone to help you make a Scriptnode compatible distressor right? Is this a 50/50 collaboration, or are you looking for someone to create the whole (or most of) the algorithm? Is it for a specific product, or just a utility you need creating? What's the project timeline?
I'd like to know the ins and outs, you've got my interest, as I already had plans to model the unit.
A while ago I set up a little python environment for prototyping and benchmarking it. With neural tools built in, machine learning for paramter tuning.
-
@griffinboy Cheers, mate - I was being more informal…I wanted to make a multi-band Distressor for some friends. It sounds like you know exactly what you're doing—impressive, and I'd love to know more about your process. Given my relationship with EL, I can provide some formal assistance, but it would need to be under the rubric of an official license (which I can certainly help with).
How well can you model amplifiers? I work with a lot of guitarists, and it might be interesting to model their amps, so they can work on the road. If you get them right, I can provide an official endorsement from them and myself.
-
Ah I see! Makes sense now!
Your network always impresses me, I don't have connections with anyone in the hardware world. I've done a fair amount of analog modelling now, but it's mostly from data sheets + schematics. I have trouble getting my hands on real units! (Student without any money)
I've programmed a collection of analog amp models in the past but I've never made a guitar amp sim specifically (I did stuff like console preamps). I could probably do it but it depends on the circuit. For example I can do circuits with tubes no problem, but good transformers are a bit tricky. I take my research seriously but I'm certainly not as clever as the folks over at NI, say. I know that some amps hit tranformers hard to get their specific character. I'd like to train an AI on transformers, same with transistors, just haven't gotten round to it.
-
@griffinboy Do you Transformers as in LLMs, or electrical transformers?
The thing with modelling is not to make an exact copy, because that's much more work than necessary. Cognitively, it's never going to sound the same to people, anyway (because sound only exists in our heads, and the context of using a plugin is different using a physical device).
It's also non-trivial to model everything—but that's OK, because all you need is small subset that models the important characteristics—users can always turn knobs, and if…say…I need to turn something to 5 in the emulation, but that would be 9 in the real hardware, that's fine — the important thing is that the emulation respond correctly when you turn the knob. In fact, it's better not to try to do everything, because then that's your focus, instead of the important bits.
And, of course, non-linearity.
Lastly, sound quality is important above all — and AI has problems there, because errors are a feature, not a bug. IK Multimedia uses amp/cabinet modelling with over 1,000 IRs to simulate nonlinearity. And it all sounds like shit compared to a single, conventional IR from Tom Syrowski.
All that being said, can you do this, assuming you had one, and could talk to the company that made it:
ENGL E658 Steve Morse Signature 20 - ENGL Amplification
ENGL Amplifiers are world renowned Guitar amplifiers for rock and metal, built to the highest quality standards in Germany. The new ENGL Steve Morse Signature 20, is a compact and versatile amplifier that packs a punch in terms of both features and sound.
ENGL Amplification (www.engl-amps.com)
If you can do this, I can get you a lot of work.
B
(Also, it's not the NI folks to worry about — it's Slate Digital.)
-
Well, I've created pretty accurate models for two of the tube types used in this amp. But not for one of them. Interesting that three different types are used!
I'd probably be able to do it, but the time it would take depends a little bit on the actual circuit topology. If I had enough data from the unit (tons of test recordings) I could probably create something in this ballpark, without having to actually tear open a unit myself. Then again, I'd probably want to model certain components such as the output transformer using AI so I'd probably end up buying components from the amp anyway. As is the problem with all analog modelling: it's easy to make an accurate model but not so simple to create a CPU friendly one. As slow as AI is, it can outperform some of the physics sims needed otherwise.
I see what you mean about 1:1 matches, but in my opinion it's not so great to create an inaccurate model of anything. If I'm going to plug together two specific pedal sims, I'd want it to sound very similar to the interaction between those pedals in real life.
In regards to the distressor, I have neither a licence for the slate digital one, or access to the real unit. But if I did (or more accurately when I do) I'll create the compressor dsp and give it to you.
-
@griffinboy You can tear open the units - I can get you any support you need from the companies. It's just a question of if you can do it. If you think you can, DM me. Or…DM me anyway, it's all good. As you're a student, I'm impressed with the work you're doing - keep it up!