Benn Veasey’s ‘Vibration’ Is a Celebration of Dance and Community Manifested by Artificial Intelligence

Filmmaker Benn Veasey’s music video for Sub Focus’ Vibration is a globe-spanning ode to the power of dance and community. Through an impressive collage of AI generated imagery, Veasey whisks us through time, space and history, taking us on a tour of humanity’s relationship with expressive movement. For his video Veasey utilised MidJourney, a generative artificial intelligence program, which allowed him to create the incredible mass of images needed to pull off his ambitious idea. DN is delighted to premiere Vibration on our pages today and is joined by Veasey for a thorough conversation where we dig into the challenges of using AI to create art, the importance of not being blinded by style over substance, and the lengthy process it takes to not only create these images but edit them together in a way that offers emotional resonance.

What software did you use for the video and how was your experience using it?

I made this video using MidJourney. It’s a tool I’d been playing around with for quite a while, probably about 18 months prior to this project. For those who don’t know, it’s a text to image generator so basically you type in what you’d like to see and MidJourney creates a picture based on that. I was absolutely mesmerised with the still images it created, so I started to play around with piecing these images together to make animations.

It needs to be attached to something with emotional or narrative weight if you want to really engage an audience.

The problem was, when it came to more photorealistic stuff the images were at best uncanny and at worst grotesque. It just couldn’t do faces or hands. You also had to lean into this really worn, grainy aesthetic, which I actually really loved, but a narrow aesthetic meant limited applications. I pitched it around a little, but it didn’t get much love so I put it on the shelf and just continued having fun playing around with the software. 

When did you begin collaborating with Las Bandas Be Brave and Sub Focus?

Vibration was really a case of the right project coming along at the right time. When the brief came through from Andy Roberts at Las Bandas Be Brave, I instantly knew that MJ was the right tool for the job. The artist was interested in AI, and by this point the software had developed so results were more consistent, and a photorealistic look was possible.

Even still, when I pitched this I hadn’t seen too many video pieces out there using MJ. Aidan Zamiri and Yuma Burgess had dropped an amazing video for FKA Twiggs, they may or may not have used MJ, I don’t know anything about the software they used, but the point is that there weren’t a lot of references to call upon for the technique so I knew I’d have to be really clear when presenting my idea. As such, I aimed to show rather than tell when I pitched for this video with a treatment that was light on copy and mainly consisted of GIFs showing the actual technique I wanted to use. 

What were you pitching from a narrative standpoint?

The top line for this video was: “…a celebration of dance and the communities that surround it, dreamt up by Artificial Intelligence”. A celebration of dance is easy enough to describe in words and still images, but a computer’s dreams? Why fumble around with prose when you can actually show them your idea? 

I always like to pitch using testers where possible: When you’re working on a technical idea, as I often am, not only does it help to communicate the technique behind the concept but it also helps to inform the creative process itself. You discover little quirks that live inside the VFX/SFX you hadn’t anticipated, and these then lead to new ideas about the characters and narrative. Making testers isn’t just about communicating ideas, it’s also a way of discovering them.

For example, in Vibration I realised that I could feed images MJ had created back into MJ, helping you to morph one shot into another. That’s a cool technique to discover, but at that stage it only really is a technique and not much else. It’s interesting but purely aesthetic. It has no heart, no character. It needs to be attached to something with emotional or narrative weight if you want to really engage an audience. So the morphing technique only really began to take shape when I applied it to faces: I discovered I could have a single, continuously evolving face made out of people from different cultural and historical contexts. It’s a really poignant visual and encapsulates the main idea I had going into the video: That there’s beauty in the sheer multitude and diversity of communities that use dance and movement as a form of expression. But I also discovered something watching all of these disparate features belonging to one face: That ultimately all of these different people are united under the single community of ‘Humanity’, and this isn’t just true across space, but also across time. Dance is as old as the stories we tell, probably older, and maybe even a way in which we told stories pre-verbally. This was such a powerful idea, the face-blending shot ended up book-ending the entire video as our opening and closing shot. But I hadn’t planned for that. So the idea didn’t lead me to using the technology in that way. It was the other way around: The technology led me to the idea.

You mentioned there about finding your way through using MJ as you went, and that project was constantly evolving, how was it communicating that experimentation with your producers and the artist?

What was interesting about this project was that we really were learning as we went along. It was a small team working on this, but an amazing one. My Executive Producer Ramy Dance from Common People Films and Andy from Las Bandas were really upfront with the artist that we’d be experimenting as we went along, and I think that was actually quite an exciting prospect for all involved. It kind of made the whole thing feel like an adventure, that we’d be exploring this digital universe as we went. Of course, we weren’t fumbling around in the dark, there was a clear direction of where we were headed, and the beats we’d take in getting there, and as a team we took care to share that vision with the client. But still, there were all of these happy accidents and discoveries along the way, like the face-blending I mentioned, that made the projects one of the most enjoyable and exciting pieces to work on.

There’s a misconception around using AI software that it shortens the time it takes to create work, or that it’s just a speedy way to make something. What was your experience with it regarding the number of hours put in?

The hours were intense of course, sixteen to eighteen hour days painstakingly generating images and animating them, as well as storyboarding and listing moments out on paper as I went along but I always had my team to check in with. When you’re head’s in something for that long, a fresh perspective is absolutely essential, from something as simple as “Let’s hold onto that character for another beat” to something as substantial as re-ordering the structure of the edit.

 With AI you’re often working with multiple variants of the specific images you’re requesting, how did you find the challenge of navigating those variants to find the right ones for your narrative?

It’s funny to get all philosophical about this software and the creative process because the actual workflow is actually quite straightforward, and arguably a little crude: At its core, this video is a simple piece of stop motion. After MJ has created an image for you, you can ask it to create variations of that same image. For example, you might have an image of people dancing in a field at sunrise. MJ will create four variants of that image where the people are wearing different clothes, they’re dancing in different poses, and the trees in the field are different. So you then create variants of those variants, and so on. The beauty is that whilst there are differences between each image, they don’t vary too drastically, they’re just different enough. Slight changes to subtle details. So this means if you put the images in sequence there’s enough similarity that you have a consistent subject and environment, but at the same time all of these intricate changing details.

What was beautiful about this process is that subtle movements or moments would begin to present themselves purely by chance: A ballerina’s arm might be raised in slightly different positions across a series of images. So by rearranging the images in the right order you’d create the illusion of the ballerina performing a pirouette. I’d then go in frame by frame, doing mini paint jobs on the images to smooth out the animation or to exaggerate it. But the key is to let the images lead the way: If you try to dial in specific details when asking MJ to create an image, it just doesn’t work. You’ll lose consistency between images, or you just won’t get what you’re after. Instead, you have to spot the movements that are already in there and lean into it. I once heard a sculptor explain that when working with wood you’re not carving your block in a way that makes it resemble the object you had in mind when you started, but rather you’re removing pieces from the block bit by bit and gradually revealing the object that’s already there at the centre. This process felt a bit like that.

The key is to let the images lead the way.

Why do you think the style of animation works so well with Sub Focus’ track?

This style of animation obviously has a certain feel and look to it, but it’s precisely because of this slightly glitchy, frenetic aesthetic that I knew it would be suited to Sub Focus’s track which has all these amazing moments of explosive energy in the soundscape. Even still, we couldn’t fall back on this style of animation for the entire video. It would be too repetitive, too abrasive. So one of the challenges was to create more gentle moments. This was essential to reflect the ethereal, more stripped back sections of the song, and to create an edit that flowed. So it’s not all glitchy animation: There’re top down landscape shots subtly animated, water rippling, dust in a desert swirling; slow push-ins on long shots for example. And most importantly, really distinct sections kept the edit engaging: We worked really hard to give each section a particular style and visual vocabulary. A slow push-in on a face and then planet earth; montage series of quick-cut top-down satellite images; ground-level wides where we watch evolution take place through time lapse; dance sequences that blend and transition into one another perfectly on the beat, etc.

Bit of a nerdy question, how did you land on the right framerate? When you’re dealing with a mass collage of images I imagine it’s important to work out the best speed to deliver that for the audience.

Really early on I decided to make most of the animation 12 fps, it just gives you time to register the motion. A lot of the AI animation I’ve seen is at high frame rates. It’s a cool look, but it’s got this hyper-real, almost time-lapse feel to it which just wasn’t the right fit for this video. I wanted the medium to be digital and computerised, but not the subject matter or tone. At its core it needed to be about people, and as such it needed to feel human, you just don’t get that when the animation feels like a hyperlapse. Also, back in the day stop motion was shot at 12fps, to save on production costs, so there was something nice about using this brand new technology but in an old school frame rate.

When you were using MJ to generate the characters featured in Vibration, was there a moment when you knew you’d stumbled across the right ones for your video? How did your relationship with those characters grow during the edit?

One of the strangest things about editing this was how familiar I got with the characters. As an editor you begin to build a relationship with the different subjects, you become intimately acquainted with their mannerisms and behaviour, subtle things you pick up on because you’re watching them repeatedly. And through this you start to get a sense of their personality. So it was really odd when that began to happen with the characters in this piece. Intellectually I knew they were fabricated people, but my brain forgot that and I couldn’t help but think of them as real. I’d be editing a dance sequence and I’d have a little internal monologue like, “That robot breakdancer really has the physical gravitas that would match the drop here”, or “That couple embracing are so tender yet sad, I hope they make it…”. I don’t know, maybe that’s a normal mental state for an animator, but this is the first proper character animation that I’ve ever done so I found it bizarre. But ultimately I think it led to a better piece, as a director, if you can understand and empathise with your characters then you’re going to get a better performance out of them. I guess that’s the same if they’re made out of flesh and blood or ones and zeros.

What are you working on now?

A couple of really exciting things actually. I’m working with a tech company to develop a brand piece that involves AI again, but it’s less frenetic than Vibration. I can’t say too much, but it’s like a meditation on the current state of technology and how it’s shaping our world. Then there’s a mixed media piece for a haircare brand, so something completely different but I love that. Variety’s the spice of life, and it’s also for a really amazing company. I know the founder and they’re a really inspirational person so it’ll be a real privilege to work together. Then finally, I’m putting together my first short film in forever. It’s how I got into filmmaking and I’m finally getting a chance to return to my roots. It’s based on this beautiful short story my mum wrote some years back. It’s called Leaf Fall and it’s set in a world where humans hibernate for three months every year. I like my speculative fiction to be grounded though, so it might sound really out there but that’s actually just the backdrop to quite a small, intimate character study.