In the past 3 days i've come across 2 companies that have analogy compressor plugins that has given me that giddy excited feeling inside, and coincidently both companies used machine learning to get there. One of them is Melda and the other is Tone Empire.

This video popped up randomly on my youtube feed an hour ago, and i clicked it instantly. Tone Empire are a company who i believe to have very underrated compressor plugins. Their OptoRed and Firechild are hidden gems imo, i discovered them only recently and i was surprised at how analog they sound. I don't see anybody really talking about them. They've got that "juicy" analog sound. They have other great plugins too like their Model5000 (loose API 2500 comp emulation), LocNess (Devil Loc), and a few others. So when i saw the video title i was already intrigued, given Tone Empire's track record.

Then i heard it. Wow. That thing is CRUSTY!

Over the years i've been a pretty regular Acustica customer (i've spent about $1,000 on their plugins). Acustica were exciting because they used a unique tech to get an analogy sound that other companies didn't do (it's a convolution/sampling sorta tech, put simply). They make amazing EQ's that sound very "alive" but they can never really nail their sampling of dynamic processing. Acustica fans don't like to hear it, but their compressor emulation kinda sucks. There are exceptions (El Rey, great comp), and it's gotten better in the past year, but most of their compressors don't sound quite right to me, especially when driven hard.

One important characteristic about using hardware is that when you drive the input knob it "heats up" the signal, and the further you drive the input the more it starts to saturate and break. Well Acustica plugins don't have that element. Like, at all. But Tone Empire have absolutely nailed it with this LVL-01 plugin, it's exciting.

***EDIT: I tested it and def liked it. Feels like more of a saturator plugin than a compressor. Compression happens by nature of saturation obviously, but i probably wouldn't typically pull this plugin up if i was looking for a compressor, it's more of a specialty use case, i'd pull it up if i want a dirty edge-of-breakup sound. Really nice for adding a smokey rich "hair" on vocals, and fun on drums making things on the edge of breakup, and blend the dry.

https://tone-empire.com/shop/lvl-01 (intro price is only 30 bucks atm too, since it just got released TODAY. btw, no i'm not sponsored lol)

Now about Melda.

I finally dived into their Complete Bundle a few days ago, did some extensive tests after years of resisting the time & energy it would require lol. They have a plugin called MTurboComp that impressed me. Really didn't expect it so it caught me by surprise. MTurboComp is like 12 analog compressors in one plugin, and they actually sound quite convincing. The saturation does that nice "edge of breakup" thing, and the compression behaviour sounds very analog to me. The more i used em the more i started to get that giddy feeling, knowing i'd discovered something that can get me the sound i want.

And surprise surprise, they used machine learning to get there. Quote from the website:

There is a quirk with the MTurboComp compression though. By default the compression is driven hard. I opened up the plugin and it started off doing -10db to -12db compression on my drums from the start, before i even touched a knob. And turning the "compression" knob down to 0% only brought the gain reduction down to -5db. So it's gain staging thing. You've gotta lower the input down from 0 to get to a starting point of -0db GR. Or you could reduce the clip gain of all your tracks. But even then you'll be getting some GR by default unless your tracks are very quiet.

Regardless, they still sound great. And it's not a huge problem if you know what you're doing.

And besides, if anything, the default GR shows off how good the compressors are, because it's rare for hardware emulation plugin to be able to compress more than -10 db and still sound great and convincing.

Anyway, just wanted to share some things. I feel like it's something audio nerds should know about.

 

Last edited: Jan 8, 2023

Tested it.

Download is as easy as can be, a direct link to the exe. But caution when installing or uninstalling, it's not protocolled and leaves some files and folders on your disk. I recommend using a tool like Revo for (forced) installation.

Seamlessly resizable GUI, from close to credit card to larger than life.

The compression meter is buggy, no way to keep the needle completely to the right.

It's a real CPU hog. In PD it took here more than a second to switch the pages, highscore, even Acustica Audio pluggies are faster!

Recovery alters release and attack.
Even at slowest setting the attack time is very short and release extremely short. At second position the release is basically zero. According to PD, not even 1ms long!

No oversampling!

Tried it on various sources. My resumee, useful with a tiny bit of reduction, distorts too fast and heavy (OTOH, you may like exactly this) and the short release 'time' renders it unusable to me.

 

Very distorty indeed. I was looking into it and while they don't say exactly what's being emulated (they just say modified STA level) because i felt something was wrong, but i think it's ment to be something like this:

This is a clone, but if you see the controls of the unit on the plugin page, it's pretty much the same, and lo and behold, it distorts like crazy. Weird but...¯\_(ツ)_/¯

exactly my thoughts. sounds cool for when you need dirt but saturation on it's own is too static. For straight up compression...Nah.

 

hell yeah they are! I have always liked their plugins and have used them on a ton of mixes. Firechild is great, as is Goliath v2 and their tape sim plugin. I think they get written off by some people that are more familiar with their Beatskillz name that may be targeted more towards beginning producers...I don't know. Whatever the reason, I heartily agree these guys are making some great plugins. I'm always interested in a new release of theirs, and didn't know about this one. Thanks so much!

 

As they say on their web site : "Designed for more saturation and less obvious compression, this plugin is perfect for your Drums, Loops, Breakbeats, Bass, and Guitars /Synths."

 

Acustica compressors are far from perfect dynamic/reaction wise.
But this one seems too. Probably due to ... learning itself.

Machine learning works well with saturation.
Compression is a totally different thing, related to average/peak/RMS ... levels.
This part is almost linear.
Better with algos/curves than any "learning".

Same with EQ : Acustica do a great job for the saturation part ... but what difference on CURVES ?
Any "good enough" EQ can do the trick. Just add saturation AFTER.

For now, i try to separate dynamic/EQ... AND saturation.
So i can use any "go to" compressor/EQ ... AND add Acustica (or any saturation) after/before it.

Yes, it MAY be nice to get both at the same time.
But you loose control of it, because to be "like the original", there is NO saturation threshold ... in fact, no saturation settings at all on most compressors/EQs ... emulations.
Ratio between dynamic compression/EQ ... AND saturation clipping is FIXED.

I will separate compression/EQ ... and saturation some more years

 

Very interesting stuff. NERD ALERT!!
I've read mixed opinions on United Plugins, of which Meldaproduction is a member or somehow associated. Haven't tried their plugins so can't comment.

I do understand what Melda did with their compressor. Actually, I wonder why is not more widely used. Well, perhaps other companies just try to sell it as "full AI". Anyways. Seems the UP guys went for the full-AI algorithm thing.

One of the most powerful applications of AI is multi-variable optimization of existing algorithms. That means you have a shitload of parameters to tweak. Imagine one hundred of knobs, sliders and faders. It's a whole science because is extremely hard.

BIAS Amp tone-matching is one good example with plugins, not using AI. Using AI you're basically optimizing all those parameters to give the most faithful response your algorithm can. Very clever. And doesn't cost CPU at all in the final product.

 

It would've been cool if he A/B'd with the analog unit.

The demo sounds nice, but I don't hear anything that blows my mind, or that I can't obtain with other plugins. But my ears are full of shit anyway

 

Funny you should say that. Happens almost the same to me: my brain is full of shit. And I'm not trying to be arrogant here...

On a more serious note, expanding of what I said: I wonder if that's what Overloud used for their already famous TH-U rigs. Just a shot in the dark, but it would make sense.

 

Are the TH-U amps that good these days? Last time i demo'd them (2014 or so) i didn't like them much, they sounded a bit digital and fizzy to me compared to other things like Amplitube. Should i give them another chance?
If so, why? What do you like about TH-U?

 

Melda Turbo things (EQ, Comp, Verb, Amp) are underrated AF. In general Meldaproduction and Voxengo are two companies that are so damn underrated providing things with so much sound quality and low CPU usage. For me it is what should be the "standard" for audio plugins. We are tired of damn analog interfaces that are fulfilled with shitty algos.

 

I don't have any interest in tone empire btw. The only software from them that i tried was buggy and so heavy on CPU that i just never think again on their products. I have El Rey from AA and i love it. But i'm not a aficcionado for the company, and i understand that most of the tools will make my job slower. But i like AA and their CPU usage is getting better by the time

 

Last edited: Jan 9, 2023

Keep in mind my guess is to make the most of the modeled components they already have. That being said...

TH-U has improved remarkably. But what sets it apart is the cpu usage versus quality ratio. Same can be said about other Overloud plugins.
Has always had a ton of amps, pedals and so on. Lately has improved in the cab section. And the famous "rigs" are entire guitar FX chains modeled from top to bottom, with few adjustable parameters. Those are even more CPU efficient and are kind of plug'n'play. It's a great concept.

Amplitube sounds great too, but it has become a Mammoth of both RAM and CPU. I haven't even tried latest versions because of that. If your PC can handle it's a solid bet.

 

So true. Hell yeah

 

I bought the SKnote STA-Limit about four years ago, but times have changed as well as my taste in how and when to infuse color on individual tracks. Now I mostly cleanly compress using the spic-and-span fircomp 2 (ridiculously inexpensive at about $28) and add harmonics and distortion after the fact when I choose to.

Since going cleaner in regard to compressors, woman are constantly hitting on me and my hair is healthier, less gray with a shine and bounce again like when I was a teenager.

Just sayin'.

 

OMG

For a moment there I thought you were gonna add "and has made me become a better person"

 

THU rig player is basically a Kemper clone
Profiling

 

I found this video couple of weeks ago, very ingenious :

 

In which way is this ingenious, please? The title itself is already wrong. A whaveshpaer is something different than a compressor and can't even replace it. It's in no way better. That genius just discovered clipping, nothing more, nothing less. Nothing extraordinary.
A clipper doesn't come with attack or release times -> inferior.
A digital compressor can be extremely clean -> superior.
And that digital compressors can't have zero attack time is simply wrong.

I don't blame you for not knowing this but this guy should do his homework before trying to teach others.

 

Ok, Sensei. I stand corrected
Tbh I never really used this plugin anyway. I just was impressed cause this guy is an amazing producer so I didn't question much his words lol.