According to KORG's website,

CMT does not simply simulate the output sound like previous modeling techniques. It leverages digital models of the components used in the analog products like transistors, capacitors, and resistors. Using CMT, we have been able to completely capture the sound and behavior of the original analog synthesizers.​

How is this different from what most (?) analog emulation developers are doing? And yet, KORG's EP and ARP 2600 sound much more 'alive' than efforts by Arturia or Cherry Audio (for instance), and even than sample-based releases, at least to my ears.

 

Korg is a small Japanese company with about 290 employees that is very innovative and hardworking, they have invented many things themselves and have quite a few patents.

Cherry Audio is in the low budget area and they get the sound they paid for for little money. A cheap sound.

Normally you can never convert hardware into software 1:1, the original always sounds better than the software copy. The Korg crew worked hard and found a technical way to significantly improve the sound and that's why they all sound better than the competition.

Arturia from France started with the Prophet V Emulation, it is based on " TAE® is "True Analog Emulation" - Arturia's exclusive technology that accurately reproduces the tone, waveshape, tuning and other detailed characteristics of an analog synthesizer.
www.arturia.com/products/technology/tae®

 

It's all buzzword bullsht.

 

Unless they show us the models and the source code we simply will not ever know. And Korg is not going to open-source their efforts. So any guess is good as any.

 

There´s also decades old Yamaha emulation technique from old digital consoles
Research and Development - VCM Technology - Yamaha Corporation

 

It's just the branded name for their proprietary technique. Even if every company did it exactly the same without knowing how the others did it exactly they'd all have different branded names. What's important is the sound, but as you've said it's better. I've overlooked some of the more recent Korg plugs, now I wanna check em out.

 

Also the people doing it beyond anyone else right now, in my opinion, is Cytomic (not synthesizers tho)- starting with his plugin The Scream and how intricate the modeling is for every component, and allowing you to tweak every component as you see fit. There's a video where he compares are a real tubescreamer to the plugin and it's just insane. I bought a different guitar pedal bc I fell in love with a preset he has where he modified the circuit in the plugin to be as it is in the pedal accurately, and when I got that pedal it's like exactly how that preset in The Scream is.

He says he's making a plugin of the SSL bus comp like that next (yes there's The Glue, but he says only the envelope follower is modeled 100% accurately, next will be every single component involved)

 

Acustica's approach is somewhat similar i believe. they use sampling / impulse response, but they also incorporate some sort of component level modeling. I could be mistaken, but im pretty sure they do.

I think this sort of approach could be really fun. Especially for things like saturation, but anything really. Imagine being able to map an env follower or lfo etc to a capacitor's value etc. Can't wait!

 

Definitely agree with your second bit there, I think even if companies have been doing this for a long awhile we're getting closer and closer to insane levels of accuracy and possibilities.

I definitely like a handful of the Acustica plugs, but I'd like to see them try to do what Cytomic did where he puts the hardware next to the software, takes signal generators, an oscilloscope, and an analyzer and shows how close the results are with the settings in the same places (and even the quirks of the gear like how it is in bypass etc) and the changes done in real time (and not once give some excuse like "well, components fluctuate..." etc etc)

 

OT but it is starting to seem like 9/10 of your posts are copy pasted from AI. Sometimes you include that you have used the AI for a summary but I think you should always do that. Maybe you didn't use it in this post, but at this point I can't tell anymore!

Anyone with an e-mail address or even a search engine can now access AI features, so it is not necessary. I do mean this as respectfully as possible, sorry if it stings.

 

Sometimes I thought that Radio was like a sort of AI bot and, as you, I mean it as respectfully as possible.
It kind of irritates me (I know that is a personal thing) because most of those answers don't go to the point but are generalizations that seem copypasted from Chatgpt answers.
What irritates me is that I think that he makes the wrong questions to chatgpt because as I said, very often, they don't get to the point.
Sorry Radio if this will hurt you.

 

RadioGPT does not have a pain processor

 

Nah, in his past incarnation he used to post links to articles and stuff as answers with the occasional human touch to it, that was before AI became a widely available thing. Lazy maybe, but not a bot.

 

AI is no different than using Google or any other search engine. If you do not create a good search string, you will get BS.

It's easy to say all someone needs is an email to ask it, but the truth is; many people are not good at Search because they do not know what to ask it. It takes about 3 seconds more than telling someone to RTFM or posting a "funny" LMGTFY link about 15 years too late to be funny.

 

I'm feeling lucky!

Here's an interesting interview that goes a bit into that approach:

 

@Radio I wonder how AI Radio station would sound?

 

Hello @nmkeraj!

It's called artificial stupidity (AS). We don't need a machine, we have an excellent brain ourselves.

A playlist can be put together quickly, you don't need to pay a software programmer.
Well, first of all, you'll fire a few people who work at the radio, the machine will take over, then they can see where they end up.

How it sounds depends on the code, the machine probably scans the charts and whoever has the most clicks and likes will be played, because after all, radio stations finance themselves exclusively through advertising. You can no longer call the radio station and talk to a person, because the machine is a machine, it only does what it is programmed to do.

Signed: Ein-Stein, head of the research department "Help, we're getting stupid"
and chairman of the association "Make a lot of money with people's stupidity"

 

Criticism, both positive and negative, is always welcome and necessary.

No, I wrote the article myself. If a user has a question and no one answers it, I sometimes use AI, such as ChatGPT and Perplexity, and write with AI. I think every question deserves an answer. Why shouldn't you use the latest AI if it leads to the goal.

If someone needs help installing or configuring Windows correctly, I sometimes copy the appropriate text that shows how to do it correctly.

It just seems that I copy 9/10 of the articles, but in reality it's max. 1/10 and less. What they don't see is that I really put a lot of effort into it and often write myself. Sometimes a few people assume that I'm a psychologist or that I googled the text.

I only use the following search engines: Presearch and Startpage.

I have created a lot of folders with text files and pictures from the past few years. If someone has a question, I look there to see if I have the answer and then post it here.

The problem here is that you don't know the people you are dealing with personally and many prejudices have no substance at all, the accusations are more felt than known and some people here are not nice people by nature and you have to deal with them whether you want to or not. Some people only think in black and white and not in shades of gray, they generalize and sweeping statements.