Hilarious Dan Worrall review of AI channel strip

[Martin Walker]

As I type this, Dan Worrall's in-depth review of the Phil Speiser 'THE STRIP 2' plug-in has only been online for 50 minutes, but it's the most depressing yet hilarious account of audio AI use I've read to date:

https://www.youtube.com/watch?v=HkWJj79HFwU

Dan does acknowledge that his 30 years of mixing experience make him an unsuitable candidate for an AI mixing 'bot', but his analysis highlights the ways that each module can sometimes make things sound worse.

Oh, and it's yet another one plug-in that doesn't offer a demo version, or a manual for that matter, although the built-in AI chatbot will attempt to answer your mixing queries if you type them in.

And there I was thinking that AI was supposed to be a time-saver.

[Matt Houghton]

I've similarly seen these marketed very heavily on Facebook... and like Dan my interest was piqued, until I discovered that there was no demo. Unlike Dan, that's where I gave up.

Hats off to Dan. It's a cracking review. "No BLEEPing boost"!

[The Elf]

Many people seem to believe in the concept that for any skill they can't be bothered to acquire there's a quick fix in their wallet. Once it was presets, and now it's 'intelligent' plug-ins. With that in mind this type of processor, and the others that will doubtless follow, will fool the unwary sufficiently to part with their money.

[Rich Hanson]

There is an interesting discussion in the comments below that video as the guy that developed it has responded - see the pinned comment (as ever with YT comments it does eventually disintegrate into mush though)

[Matt Houghton]

Many people seem to believe in the concept that for any skill they can't be bothered to acquire there's a quick fix in their wallet. Once it was presets, and now it's 'intelligent' plug-ins. With that in mind this type of processor, and the others that will doubtless follow, will fool the unwary sufficiently to part with their money.

To be fair, there are some miraculous 'quick fix' plug-ins around now, and I don't think it unreasonable to look to this sort of technology to achieve things. Some save time. Others do things that weren't previously possible. Others do a quick job inexpensively, albeit imperfectly. As Dan says, just because we have invested years in acquiring skills, knowledge and experience it doesn't mean we have a right to gate-keep...

But I don't think Dan was criticising this one for offering a quick fix; he was criticising it for the design, execution and the sales/marketing tactics. And I'd say his skills, knowledge and experience put him in a very good place to judge.

[Martin Walker]

There is an interesting discussion in the comments below that video as the guy that developed it has responded - see the pinned comment (as ever with YT comments it does eventually disintegrate into mush though)

Wow - that discussion has really kicked off, although to his credit Phil Speiser has entered into the spirit of things with a lot of his own detailed thoughts, as well as remaining polite throughout.

[SafeandSound Mastering]

Many people seem to believe in the concept that for any skill they can't be bothered to acquire there's a quick fix in their wallet. Once it was presets, and now it's 'intelligent' plug-ins. With that in mind this type of processor, and the others that will doubtless follow, will fool the unwary sufficiently to part with their money.

As Dan says, just because we have invested years in acquiring skills, knowledge and experience it doesn't mean we have a right to gate-keep...

The only problem being is the statement is grossly premature and potentially totally wrong. It may never become as good as skilled humans with ears and emotional intelligence. Numbers are not the be all and end all of intelligence.

That is worth considering as it may be a reality that it can never surpass human skill. It is not written.

It is important not to state possibilities as foregone conclusions.

When the results are as good as a very highly skilled human who is really listening, with a very extensive and varied high quality tool set, then maybe.

That is going to be a task that I personally would not wish to try and embark on. And I would say I in a very good position to know.

I am about to purchase another EQ, I must have 20 in use for my work, I think nothing of it, if it adds something unique to what I can do with sound. I am extremely doubtful such automatons will be so discerning in what tool to use for a specific task.

These automatic processes fall so very deeply short, it's borderline ridiculous they are even entertained beyond curiosity.

It will also depend if you are comfortable with automatic processing ( an automaton) changing aspects of your creative output that you may never understand. And in addition never learning anything about why something does not sound right. Which a highly skilled, listening, human being is likely to be able to explain to you.

[Dave Rowles]

AI can only recreate something that's gone before. It can give you options that you may not have thought of perhaps, but it's all based on what it's trained on.

There are some studio junior/trainee tasks that can be off loaded onto AI, but I've not heard an "AI mix" that doesn't sound bland and uninteresting.

[SafeandSound Mastering]

What actual Ay aye is a little confusing we know there is a technically strict definition. And they are throwing it all sorts of tasks even within just the audio world.

And have done for an entire decade. I am still here, I am still working So in 10 years thus far, to my eyes, it's failed. I can only speak for myself. The reason is, it has not met the grade, not good enough.

The elephant in the room is the excitement and hype is unwarranted relative to result. I think some people want to the on the cool train, they are free to do so. (Personally I find it a massive bore, it's yawnsville to me compared to my actual work and enjoying making my own music.)

That is irreplaceable to me.

Beyond the self indulgment (not always of course) there is a process that has a therapeutic value (and basic sheer enjoyment) in making music which is evidently common amongst the very vast majority of people who make music.

Diluting that seems self defeating and short sighted.

I am happy ay aye is able to find cancers quicker etc. Humans should be extremely cautious about where they let it encroach in their own lives. You know tech... once your in... it can be difficult to back track. Myself, I find it a massive bore, my cat is so much more interesting that ay aye. Ay aye is not life. To me it is not even interesting, I read paper books they are great, filled with actual intelligence ! (some)

I am no Luddite but I can reason and ay aye is going to be very much a secondary burn, if that. I have better things to focus on in life (like... literally anything !).

Your existence is based on a theorised 13.8 billion years of evolved, cellular DNA intelligence.

How about we reflect on that a little more.

That's merely one view of course, my view I am sure it is just one of many.

[Matt Houghton]

As Dan says, just because we have invested years in acquiring skills, knowledge and experience it doesn't mean we have a right to gate-keep...

The only problem being is the statement is grossly premature and potentially totally wrong...

...and possibly misunderstood too I don't *think* we're really disagreeing on anything here.

To me, gate-keeping means someone with a vested interest controlling or denying access to a market. I don't think anything you said suggests you or any of us are doing that? These things are just tools. As with any tools, we need to assess what they do, forming judgements based on our skills and experience. Sharing honest opinions based on such assessments to educate people — or even to emphasise our own worth — is not gate keeping.

Don't get me wrong. The whole AI marketing-hype bandwagon irritates me almost beyond measure. It's just a buzzword, really, and we're a long way off anything genuinely 'intelligent' reaching us. But while nobody should ever buy a plug-in because "it's AI so it must be good," equally, people shouldn't have to avoid those tools purely because of what a marketing team writes. If they appear promise something that's useful to you, it's probably worth trying them and forming a judgement.

Yes, there are plenty of 'AI' tools that fall short of what a human can achieve. There are plenty that are, frankly, crap. And there are plenty where the marketing blurb massively overpromises. But you could well say the same of any number of other specific tools over the years.

But there are genuinely good tools too. I use several machine-learning based plug-ins for certain tasks, and they enable me either to achieve things I couldn't manually, or to achieve things in a workable time frame, and the time required to attempt it manually would be prohibitive. That's not just potential; it's genuine benefits here and now.

The biggest problem comes when users at whom a tool is being targeted lack the skills and experience necessary to judge the output, or the ability and tools necessary to act on that judgment where they do notice problems. There are probably greater risks with AI tools than others here, partly because some attempt to achieve so much at once, and partly because the nature of the technology can often make it difficult to refine the result... making them not dissimilar to presets.

At present, I'd say that's probably a more obvious problem in some areas — eg. 'automatic mastering', given that many of those who might feel the need for automatic mastering won't be able to hear/notice/correct what a good mastering engineer can, and the tool certainly can't give the same feedback on a mix that a good ME could. In other areas, the results might well be more useful — eg. dialogue noise-reduction and reverb removal, which isn't yet perfect, but is already streets ahead of what anyone can do manually without ninja-level skills and a week or two spare in which to do it.

[RichardT]

I am quite optimistic about AI in the longer term, as an assistant. We’re now in a bit of a dotcom boom, but when the inevitable disillusionment passes, we’re going to have some very useful tools.

When it comes to creating stuff, though, the picture is more murky. I think we could easily have entire movies created by just a few people, for example. Starting with a storyboard, it might be possible to create everything automatically - including images, audio and music. The economic impact would be huge. But it’s not clear whether people would want to see that kind of movie.

[SafeandSound Mastering]

Yes the definition to me is similar, someone or an entity who allows or denies access. No one is doing that on this post. I don't think anyone can be a gatekeeper. Ay aye has been given a free hand in all areas of development. And in some countries more than others.

So far there is nothing worth gate keeping against. Except maybe very much over the top hype.

For myself I do not find the "product" on sale interesting on any level really. Challenge has a value in my world, easy answers, that might very well be wrong (egg meet face), incongrous or that fall way short of what real value is, is a waste of my time.

I want to learn about music myself and have happy accidents and discover. I don't want my music making process to have watered down discovery value. That is one of the most enjoyable parts.

Wondering what happens if you do this or that or just being in the moment moving your hands around on what can be vaguely called chords, in my case and the excited rush to remember and record them. I may be lucky in having a small degree of vision when I start a track. I find this is only caused through having natural and unfixed breaks between music. I never force anything and just let the tracks come when it feels ready to. I do not stress myself to make something for the sake of filling a few hours on a Saturday afternoon. I wait, then when it feels right, I am ready to go and am deeply into the process.

Work out really quickly if tools are taking something away from you is my suggested advice.

Suffice to say a proficient mix appraisal before the mastering job can be of very high value to some. There are so many things auto DSP cannot do, fortunately. It's why human mastering (by those who are very good at it), still has a good standing in the audio world, for musicians, producers and labels at all levels of ability.

[ajay_m]

I would kill for an AI-based solution that can do 'fit to improvisation' without all that faffing around putting markers in.
i.e I play a piece of music without a metronome into the DAW and the bar markers obviously don't align. Then any human can easily make out the beat and although some bands use click tracks, for sure, nobody would ever suggest a click track was necessary to get a group of musicians in sync.

Now I want a music score based on the MIDI I played. To do this I want the AI to figure out the beat and align the bars - something a human being can do almost effortlessly when listening to music, just tap along with the beat. But to get my DAW to do this I have to laboriously go back and indeed record a 'click track' so the DAW can figure out the bar lines. This to me is where AI could surely do this really easily, it's astonishing that as far as I know, there's no easy solution, because every single musician who wants to create a score from something they play into the computer would for sure love this capability. Otherwise you have rigid metronomic music with no tempo variation.

Now I know this is common in modern music - one afternoon I recall being stuck at some traffic lights and to my astonishment the music on the radio sync'd up exactly to a flashing light on a lorry next to me - and it stayed in sync for over two minutes. Because of course with crystal oscillators and all that, stuff can be very precise time-wise. But real music breathes. It does NOT play at strict tempo. Of course, stuff like the Akai Force doesn't believe there are even time signatures other than common time. So playing something where time signatures vary in a single piece is even more of a faff, yet, this is fundamental to music scoring to be able to easily handle this sort of thing. Sigh....

[Sam Spoons]

I would love that, I hate click tracks with a vengeance, music needs to breathe. My Boss Loopers can calculate a tempo from the loop and correct a slightly mistimed 'end loop' action and my DittoJam loopers can sync to the drummer so that when we push or pull the tempo in an improvisation the loop tracks the changes in real time so generating a tempo track from live audio should not be too much of a step further.

[SafeandSound Mastering]

I would kill for an AI-based solution that can do 'fit to improvisation' without all that faffing around putting markers in.
i.e I play a piece of music without a metronome into the DAW and the bar markers obviously don't align. Then any human can easily make out the beat and although some bands use click tracks, for sure, nobody would ever suggest a click track was necessary to get a group of musicians in sync.

Now I want a music score based on the MIDI I played. To do this I want the AI to figure out the beat and align the bars - something a human being can do almost effortlessly when listening to music, just tap along with the beat. But to get my DAW to do this I have to laboriously go back and indeed record a 'click track' so the DAW can figure out the bar lines. This to me is where AI could surely do this really easily, it's astonishing that as far as I know, there's no easy solution, because every single musician who wants to create a score from something they play into the computer would for sure love this capability. Otherwise you have rigid metronomic music with no tempo variation.

Now I know this is common in modern music - one afternoon I recall being stuck at some traffic lights and to my astonishment the music on the radio sync'd up exactly to a flashing light on a lorry next to me - and it stayed in sync for over two minutes. Because of course with crystal oscillators and all that, stuff can be very precise time-wise. But real music breathes. It does NOT play at strict tempo. Of course, stuff like the Akai Force doesn't believe there are even time signatures other than common time. So playing something where time signatures vary in a single piece is even more of a faff, yet, this is fundamental to music scoring to be able to easily handle this sort of thing. Sigh....

What a novel idea. Using ay aye for something audio/music that you actually do not want to do, (or feel uncomfortable taking the credit for) that is actually beneficial to humans, tiresome for us, saves time and is difficult.

It'll be a while, as it requires actual intelligence.

[Sam Spoons]

[SafeandSound Mastering]

Something I could do with, possibly classes at cheating but you know it would not even matter if it was not perfect and ended up sounding a little auto-tune-y.

If I played a guitar solo for say 8 or 16 bars a plug in could be used in off-line mode post recording/take, so for Cubase select the part once trimmed to the precise bars length. Right click and select such a plug in from menu...

Then maybe add a few optional parameters (scale/key), length tempo (probably easily detected) and make it sound like a good solo.Timing wise and iron out to sloppiness.

So it could compare scale, drop your duff notes (fill in with suggestion notes), tighten it up to a internal grid a bit edit/tiny crossfades, you know 95-97pct quantized, do some cross fades to smooth things out, and maybe spit out a few options you could comp together and drag and drop into your arrangement.

I should clearly be working in this field ?

You could call it Afterfret, or maybe Fretgrid, Solobuddy ?. I am only partly serious of course, for now I will probably just do my many takes.

That could be built into a DAW stock.

[ManFromGlass]

You’ve literally described my process in Logic. Logic has some new tools to make it go quicker but it’s the same idea.
.
The is another option - learn to play better and practice every day but ——- naw! The life clock is ticking and I can get my ideas close enough for the digital massage into a masterpiece (har har!) Ok something that I like the sound of.

[angel4777]

i should start using ai to write ai songs lol