Every few years there is a digital goldrush, a business buzzword that makes front covers and clickbait headlines that ‘THIS WILL BE THE END OF X!’.A few years ago it was blockchain and NFT that was going to ‘change everything’ – your neighbourhood bearded Cassandra knew it would not work and the hype bubble crashed dramatically last year *popcorn.gif*. After all that spamming from crypto vros, that was delicious schadenfreude.
Now it’s the turn of AI – Artificial Intelligence, which is really machine learning, and it’s showing real promise – as well as problems – in the worlds of art and music – and I think podcasting and the personal is next.
AI has already crept into making images from prompts (I am not sure people have spotted that nearly all the original imagery on this blog and also my mashup covers is now generated or processed in some form by AI. Why? Well having had aggro and shakedowns from photographers in the past for non-profit fair use topical uses of their work, it’s just a relief to use something that pretty much guarantees (under the current legal framework) to not have it’s grubby hand-out in future.
I’ve even had copybot approaches from artists who made their career via viral work, posted online – and then they 8 years later want money for the free promo I gave them. The cheek.
It’s strange, I’ve had far less aggro from music labels than photographers…so I am for one am not crying for the people who probably will be put out of work by AI image prompt generation like Midjourney, Night Cafe, PlaygroundAI and DALL-E 2.
The stock artists and photographers churning out endless images of women laughing at salad and generic ‘funky’ illustrations saying HAPPY 2023 and LIVE LAUGH LOVE, the artworkers churning out endless mockup variants, the product photographers endlessly re-photographing thousands of pieces of stock at various angles…these drudgy repetitive jobs should have gone the way of the Dodo aeons ago, and let the creatives to be free to actually create.
Artists will Work It, Flip It And Reverse It
I say this as someone who used to work as a professional designer and works as an artist currently. Sure I have big reservations when my work was stolen for AI training I wasn’t paid, and I think there’s a lot to work out re: royalty payments and I think keywords around living artists and photographers should be blocked by the technology, or at least royalties paid. I refuse to use those prompts as matter as principle – there’s thousands of years of work to choose from that doesn’t need to parasite a living artist AND looks way better. Stop being basic af copying others and learn some art/photo history.
And yes, I think deskilling artists to get people back to office jobs is not what we should be aiming at here – the idea isn’t to get Artificial Intelligence to replicate badly and generically what’s gone before like some content soup, stir and remix and voila – I don’t think that would fly anyway, since design by committee never works. Machine learning has learnt by everything that has gone before, even the bad mistakes, you have to tell it like a particularly stupid child, breaking down each step and banning it from stuff it likes and gets obsessed over (grids, symmetry and order, seemingly) This can actually be helpful for your own process – why do you do what you do? What are your influences? How do you do that thing?
We want Luxury Fully-Automated Gay Space Communism, and AI might just be one way to get it.
(Although I still think those creating Manic Pixie Dream Girls with Greg Rutkowksi prompts should get a life, since although I agree with him over AI cloning living artists, his work isn’t even that good to begin with! Copy someone older and better, like the Goya he’s desperately fanboying please? Also these dodgy cishet male fantasy soft core images of women are all kinds of…eww. Just stop. I’ve been countering these by making my kind of erotica – photos of not very clothed bearded old men, fuck those beauty standards :-P).
So with those moral qualms aside, I do feel that ignoring AI is like saying ‘photography will never catch on’ in the 1870’s. It’s here, and the genie is most definitely out of the box. I don’t think artists and creatives will ever be made extinct by this technology, but they need to adapt and hack it for their own uses.
Art with Me, Yeah, Slow
There will always be a space for old-fashioned artwork, I don’t think AI is ever going to stop oil painting or watercolour being a thing, it struggles to look 100% real if you know those mediums well, and indeed as it’s copying what has gone before, it’s a real art to make it look interesting and not generici, it’s not ‘press a button and go’. You have to have imagination and creativity still, and a very deep knowledge of what you want the AI to create. Hence why people are copying prompts left and right – I’ve been using my art knowledge and studio photography training in my images, and it’s still hit and miss many times. Infinite monkeys and all that.
Welcome to the Church of What’s Happening Now
So how does this affect music and podcasting? Well AI assistants are popping up everywhere – from iZotope Ozone’s mastering assistant which I use almost daily, to chatGPT and other plugins being used to write the SEO blurbs on this very blog and marketing bumf. Soon we will have the ability to say something like ‘Hey Alexa, write me a blog about this subject’ or “Hey Siri suggest some topics around this subject’. And as the below video posits, you could soon have AI music generation by textual or verbal prompts, I suspect sooner rather than later. And I for one welcome our robot overlords over these (and blowing apart the barrier to entry for music creation).
In fact we sort of already have a less flexible version in sites like Beatoven.ai, but the results are far from musical after having a play…I got something that sounded more like an experimental slightly atonal sound piece than the electronic restless or joyful bed I was expecting! Also it crashed and gave errors several times, refused to play the track eventually, all the fun you get from browser based systems — partly why I don’t really want to pay money for them.
We have AI chatbots, AI keywords and description generation, and my favourite currently Whisper AI – free AI transcriptions which are pretty accurate, amazingly so and I use them on my current podcasts. Spreaker have a version that runs in a browser – unlike the local version, I found it a little buggy for longer podcasts like mine, but early days. Both versions hate long gaps and start transcribing lyrics, so if you are a music podcaster, you might want to create a vocal-only track with few gaps between sections as I do.
Adobe have rolled out some tools for turning badly recorded audio into ‘studio quality’ – in my tests, it actually didn’t do better than my vocal chain, but I think the use case is more for those with laptop mics and Zoom calls. Pretty good though, for free. Thanks to iWillBattle for that tip, but I really don’t trust Adobe to turn it into another subscription model…and we will see a lot more of these tools shortly.
Don’t Enjoy The Silence
AI silence removal is another area I’ve been looking at; because that’s a chore, and it’s hard to make it sound natural. Again there are many online subscription based services, or ones tied to an online DAW or edit system nothing local yet as far as I know. If you know of one, let me know in the comments! (please don’t plug some online subscription systems, I prefer stable solutions that won’t disappear overnight – i.e. my own computer, and one single payment, thanks).
And we of course have stem separation, which in the last year has revolutionised mashups and remixing – from the stuff that’s built in to RX 10, which usually isn’t as good as the likes of MVSep and Lala.ai – which are free or cheaper. When it comes to mashups, the brave new world of having stems for any track is pretty much here.
At the moment, well, chatGPT writes terrible code and gaslights you about it working (yes I tried to get it to write some PHP for this blog, did not go well), and lies about you if asked search engine style queries (apparently I am called Martin Tynsky and did one of GHP’s mashups?!), and BingAI may be more accurate, but is like Marvin The Paranoid Android with ‘Genuine People Personalities’ – it gets rather down if you ask it too many questions!
Personal JAI-sus
But I think the future of AI will happen when it becomes personal – when it learns not just from the wider world, but from the local creator. Some might fear this as being really dystopian, the idea of handing over your voice, style, self to a machine, and it can be used in bad ways and probably will.
But something I’ve always wanted to do is be able to edit the podcast without doing re-records and/or just type my podcast and have it spoken in my voice. The voice cloning tech is pretty much here, from Microsoft’s recent offering to ElevenLabs although it’s expensive (remember any amount is expensive to a poor podcaster like me, but you’ll quickly find that the free/cheap tiers don’t include custom voices).
I am always sceptical of any of this tech if there isn’t a ‘free trial’ cos past forays have found that it sounds terrible. There is a free offering Tortoise TTS which I’ve been using, but it still makes me sound American!
Another dream is that this is all built into a DAW, as a plugin or some-such, and I can edit via text rather than faffing around with traditional editing. Again, this is sort of here – online video editors already allow for editing by text like Type Studio but again, I always prefer a local version I own and not somebody else’s computer (aka the cloud). I do think that will happen, we will see either new DAWs around this stuff, or the existing ones will adapt.
REAPER should have Whisper support and AI silence removal built in stat, really.
I’ve seen the future and it works
So here’s to a brave new world of AI – one that could be dystopian, but used correctly and wisely, could actually free artists, musicians and podcasters to create rather than spend their time doing the boring stuff.
Time will tell which one it will be….but like those infinite monkeys, don’t expect it to write you the next Shakespeare unless you are happy with spending a long time explaining what you want, and putting your (human) originality into the ghost inside that machine.
Leave a Comment! Be nice….