Machine learning meets creative content

Published on
March 30, 2021
Chandar Lal
No items found.
Machine learning meets creative content

It might sound counter-intuitive, but the next evolution of creative content is coming from machines. Thanks to advancements in machine intelligence, a new breed of creative tools are  coming to the market. Not only do they make content production easier, they take the drag out of the creative process itself. But will they replace human genius altogether, and can synthetic content ever compete with the real deal? In our latest Mosaic Patterns Clubhouse, we chatted with Thomas Walther(founder of TapeIt), Victor Riparbelli(co-founder of Synthesia), and Nathan Benaich(founder of Air Street Capital) to find out.

Machine Learning tools are reinventing the creative process

There are two types of creative tools. On the one hand, you have tools that enable and democratise content production. They’ve existed for years – things like synthesisers, Photoshop, even word processors. On the other, you have tools that help you be more creative yourself. They’re newer, often unseen and for Thomas, much more interesting. Take the camera on your smartphone. “There’s so much going on that you don’t see”, he explains. “Knowing what scene it is, different levels of brightness, light balance – it all helps you be more creative and focus”.

A new generation of machine learning-enabled tools are coming to market that help reduce the creative friction of building something, rather than just the process friction. As Victor sees it, the next decade will be defined by the merging of the two. Video production will extend beyond cameras and editing software to tools that can take a piece of text, analyse and summarise it, turn that text to speech and create a 30-second summary video. It’s taking templates and filters which make some of the creative decisions for you (like those you already offered by Keynote, Instagram etc), and using machine intelligence to turbocharge them.

Creative ‘busy work’ will become invisible

As Benedict points out, “a computer should never ask you a question that it could work out the answer to itself.” Over time, we’ve worked on filtering those questions out – things like which printer you have, where you want to save a file, what cable you’re using to connect. They’re noisy and unnecessary. Computational photography has followed the same trajectory. A decade ago, there was an HDR button – today, a smartphone camera it will shoot five frames and save the best one, without you having to ask it to.  

More creative tools are coming to the fore to take what Benedict terms “the busy work” out of the creative process.  Machine Learning can now remove the ‘ums’ out of audio, or remove coughs from recordings of live performances. They feel like rocket science now, but thanks to the pace of development, they’ll soon become invisible, too. The direction of travel is away from endless configuration buttons on creative software slates, and towards intelligent templates that enable goal-oriented creation and design.

Technology enables new use cases, workflows and ideas

For Nathan, the most promising innovations are those that are less fixated on the technology itself, and have a more holistic view of user problems and potential implementation. Founders who have “intimacy with creative problems and know how to rapidly wrap machine intelligence around them” to demonstrate demand are preferable to those who start with a cool demo and then have to figure out who to sell to.

It’s a common mistake. Synthetic media companies are often too product-focused, and end up building tools users don’t really care about. “Technology should be used as an enabler to create new workflows rather than slotting into existing podcast editing apps” explains Victor. Machine Learning offers a fundamentally new way to produce media content, and that’s why it’s exciting. Synthesia started out attempting to make video production easier and faster, but soon realised it was focused on the wrong problem. Instead of replacing video, it could unlock all the text-based content customers would never have made video for, and create a much richer experience. 

Today’s creative tools will go the way of the digital camera

Many of the tools being developed today could evolve into completely new types of media and content experiences. The next generation of video and audio tools won't just add new features, but enable fundamentally different ways of creating content. 

Thomas expects them to replace today’s technology in the same way that smartphones replaced digital cameras. “Ten years ago, digital cameras took up most of the space in consumer technology stores. Now, they don’t exist”. That’s because they actually complicated the user experience, in the same way that a lot of creative tools do for music production today. To record audio, musicians need an interface, mic, mic stand, cables, software… when all they really want to do is play the piano. The big revolution will be in dispensing with the unnecessary elements, and making other modes of creativity as easy as taking a picture on a smartphone is today.

Benedict agrees. The evolution of point and shoot cameras in the 1970s through to digital cameras, SLRs and finally smartphones show us what we can expect from creative tools: “everything will be software”.

Technology doesn’t replace genius – it liberates it

But that doesn’t mean that we can say goodbye to ‘real’ audio and video, created in the way and using the hardware tools currently familiar to us.

It’s a mistake to think that machine learning replaces human genius – it just removes the tedious parts of the creative process, and all the micro decisions that create drag. Intelligent tools allow humans to abstract the process up a layer, letting them bring ideas to life seamlessly, rather than getting stuck in the weeds of execution.

The ultimate vision, says Victor, would be to create a Hollywood film on a laptop. Synthetic media that replaces cameras with code is the next evolution of videography. Once it’s software it’s infinitely scalable, accessible and comes with no modular cost, – meaning creators will be able to make “all kinds of weird things” that we haven’t even thought of yet.

Ultimately, these will be new genres of content – not replacements for the ones we already have. It’s about giving creators and consumers choice. Making a virtual model may or may not be as good as the real thing, but that isn’t the point. “Drum machines don’t sound like drums, but they still sound great”, points out Nathan. The value of new creative tools is that they help us to create new things, not just do old things better. 

Thanks to the intersection of machine learning and creative tools, there are a lot of exciting ‘new things’ on the horizon. If you’re building in the space, please feel free to get in touch.