LANDR etc – Do You Want To Sound Average?

Checking your mixes & masters on systems you know well, at your own pace, is ideal. They’re optimised to sound best possible on broadcast, clubs systems and Spotify, vinyl or cassette,  So that aspect of LANDR and automated mastering is good. But your vibe matters – and your reference points. So with a mastering engineer you should mention it, however vague the feeling. A huge chunk of this job is to translate non-technical listening into real sonics. Touchy-feely, infinitely variable, shape-shifting expertise is a no-go area for even the best Elon Musk-esque AI coders. There are soooo many ways to take a track, from precision clarity to slammed analogue clipping.

We need vision. Not one size-fits-most software solutions.

Let’s look at why LANDR and other automated systems and presets for mastering all-in-ones are destined to soar to the heights of mediocrity.

Distorted Reality

Mastering and mixing is not about some pseudo-scientific idea of flat frequency response and clarity of each instrument. Quite the opposite. It’s a truly visceral endeavour to create a warped reality between 2 small speakers that tries to emulate the buzz you get at a 120dB live gig – or an intimate acoustic recital. Hyper-reality is 100% necessary.

We don’t usually like stark HD pics under harsh fluorescent light. We like grain, blurring, contrast and soft light. Yes, we prefer ‘analogue’ distortion over clinical digital reality,  but only certain types of distortion resonate with the intended emotional grab. Each track is specific. Sometimes you want the juxtaposition of crisp edges and pushed tape warmth. So you must know your even-order harmonic distortion from your odd-order. One warms, one adds edge. Then there’s enharmonic distortion, which is perceived as aggressive. What about the phase distortion of legendary analogue EQ Vs linear phase transparent mastering tilt? Totally different beasts.

LANDR doesn’t understand your intentions.

Emperor’s New Clothes

LANDR couldn’t care less how you want it to sound. I  learned the hard way as an artist how little even some mastering engineers know or care.  The 90s big-name Soho based mastering wizard didn’t deign to reference the huge tracks that we’d referenced at mixdown. You see, he had ‘golden ears’ and apparently the monitors were so superior to those that mere mortals could afford, that my input was just not perceptive enough.

Fact is, the emperor was not dressed in the finest sheer garments that only the suitably aurally refined could see – he was bollock naked.

The masters sounded crap. I remastered them a year or two later when I’d pulled the wizard’s cloak and figured out how to master as well as I could mix –  by referencing … and sheer determination. The inter-relationship of mixing and mastering can be mind boggling, even to the experienced. Even with talent and application, it’s a long, hard learning curve.

Be Truly Critical – The Good, Bad & Ugly

Can LANDR type systems advise you on which elements in your mix are gonna disappear in mono club systems if we don’t sort out this mix? How good are you ears to check the results? Can you tell if it’s likely to cut well to vinyl, or sound great on that huge club rig … or iPhone?

If you want a cringey example of the kind of uncritical skills that many will be fooled by, listen to head honcho Steve Slate proudly slicing the bottom-end punch off the mix  in his hilarious demo – listen  to it on some decent full-range monitors or cans. Hear and feel the really low punch disappear with the Mystical FG-X at the settings he chooses. It’s not (necessarily) the kit – it’s the skilled and creative deployment, and the critical listening that counts.

That’s the head of a good plugin company, not your mate Trev who’s got some cracked software and a knack for Instagram. You can do a lot of damage to tracks with a little knowledge, and never hear it until it’s too late to pull them.

Mastering, Myth BustedKung-Fu Panda blank scroll. Can LANDR match the mastery of an experienced mastering engineer?

Price is no guide, don’t pay through the nose, don’t expect miracles from cheap or free services like LANDR … and remember, those who can master and mix don’t spend their time trying to get ad revenue on YouTube.

Truth is, just like Kung-Fu Panda says, there really is no magic secret to mastering. It’s always a journey not a destination. Doing it consistently well takes super-honed hearing, Zen discipline & years of experience. Mostly it takes time to learn the true level of humility you need to not let your super obedient and compensatory hearing be fooled by your own ego.

Try 12dB (careful with the speakers) of 5K shelf EQ on your favourite current hit mp3 dropped into your DAW. Nasty.. right? Now notice what happens when you bypass the EQ. … Woah! It now sounds duller than it started 3 seconds ago! That’s truly profound. Your hearing, like mine and  everyone else’s, is not golden – it’s deeply flawed and unfixable. It’s constructed with the help of our  visual cortex and our imagination – and it’s relative. Even though some people’s staggering musical pitch perception is absolute.

(Over)Active Hearing

Any pro will tell you stories of musicians who can tell you exactly what notes are in a complex chord, and which are slightly off and which way – and about the ages they dimly spent EQing stuff, accidentally using the pots on the blank channel next to it. The unsettling thing is not your own idiocy. It’s the fact that your hearing was making the change your fingers were dialling  – just not quite enough!!! And on top of all that weirdness, our hearing is still infinitely more sensitive to musical nuance than the best metering and measurement we have. It had to be that acute – for our survival. You need to know intuitively where those pesky tigers and sheer drops are in the dark.

Gear Hype

Mastering definitely needs top-end kit and great full range monitors. There is just so much ill-informed, ungrounded opinion out there , and pure marketing hype.  Price is not the deciding factor. For example, for several years  I’ve hosted a double-blind vocal mic shootout at De Montfort University.

The 3rd year undergrads consistently choose the AT2020 in the top 3 or 4. That’s against the C12VR, M149, U87, C414, TLM102 and many more. Some unjustifiable price tags outdone in practice by an 80 quid mic!

The other ones that go in those top slots regularly are the humble SM58 (£79), Coles 4038 (£800) , Sontronics Omega (discontinued but £300 ish), EV RE20 (£450) and the Rode NTK (£350). Personally, I love all those, and the Aston Origin (£250 ish). It’s not in those tests but it sounds amazing with our Drawmer 1960 variable-mu tube compressor and pre. Stunning … and not just for the price.

Extremely Costly Ported-Speaker Bass Reverb

Speakers and powered monitors are even worse – mostly bought by ‘prosumers’ on pure marketing – completely untried.. I choose ATC monitors alongside custom built, tri-amped Celestion Ditton 25s. Both of which are fully sealed. I don’t like the bass reverb from ported cabs – any of them – whoever wrote which paper about phase, time alignment and waveguides. It’s all in the waterfall plots. Good ones drop off a cliff. Damping is the hardest thing to get right. It’s also essential that you have big main drivers to get down into the bottom octaves properly without a separate sub, preferably 12 inch or more. The wavefronts must be coherent. The mix lives or dies in the six-ish octaves below 1K. Most modern monitors force your ears into compression at high frequency and don’t have the grunt underneath to vector the shockwave accurately. There are some seriously big names and silly price tags in that camp – great on paper but meh in practice.

AI – What Is It Good For?

So maybe you still don’t think it’s worth paying for expertise and kit? Good luck with the automated services like LANDR. They can’t & won’t be able to connect with your music. I studied Artificial Intelligence (AI) at PhD. The AI software ‘self learns’ statistically from analysing big datasets – of other tracks. Useful maybe for underwriting health insurance, but insufficient here in the sense  that it will never learn how to construct a mix from that data. Just like no one will ever learn how to write a great novel from averaging the words and order of every Times number one best-seller. Though that average might be worth studying along the way…

Sound Average?

Automated mastering like LANDR implicitly assumes there’s some sort of optimal sonic profile for a genre that’s  measurable. It’s simply flawed from the outset. It comes from the same mindset as those who think mixing and mastering are just some colour-by-numbers exercise in delivering their opus in the logically obvious zeros and ones.

With automated mastering there are similarities in loudness etc, true, but it’s like trying to improve the genius of Picasso’s delivery with a self learning inkjet.  So it can sound OK, sure … but it’s soulless and samey. Because the tendency is to make it sound like other tracks! That’s not our goal in mastering – it’s 180 degrees wrong.  It ignores the unique sonic profile and imposes a genre average, so it’s even less likely to stand out from the crowd – but it may well be as loud and as bassey.

LANDR – Apparently Intelligent?

LANDR has a lot of solid research behind it from the team at Queen Mary’s, but as far as I can see (and hear), all current AI software is really bad at tangential ‘lightbulb’ thinking. The stuff we need to be artists. The only path I’ve seen show promise  is where the AI is in a physical humanoid robot, and the robot must undergo an effective ‘childhood’ interacting with the real world. Gravity in particular seems to be key for the robot to start to transpose concepts from situation to dissimilar problem-solving situation – possibly a basis  of ‘out of the box’ thinking. And it’s incredibly formative research. So don’t hold you breath.

Don’t Panic!

AI big-data learning engines will never distil the essence of ephemeral beauty that an experienced mastering engineer will channel.  Automated services will only consistently blind-beat a good pro when someone figures out the stochastic hill-climbing algorithms for the meaning of life. Cos as Thomas Hardy nearly said “computers won’t self-learn that sometimes beauty lays not in the thing, but in what the thing symbolises any time soon’.

AI is amazing at big data sets and self driving cars that decide to run old folks over rather than kids  in a crash dilemma (huh?).  But as far as compassion, love and high art go, we’re still in the dark ages technologically.

Heres the mantra in Nottingham yout’ -speak :

Nah  duckeh. Not forteh two. Dem black-op mans Qbit computah bear siiick, but dem Slartibartfast mice ting bettah innit.Smile, LANDR is only Apparently Intelligent