Thoto × A$AP Rocky Tour

From there the three of us, me and two of my close friends, got the Friends Group Chat going at 5:44 PM EST that evening to brainstorm, and that's where the creative direction really got built. Friend-2 is a good friend of mine, someone I've met and know personally, and they came on early as a real asset on the creative team, deep in this tour and the vibe. Fast and totally no-ego, I'd throw something out, they'd react in real time, and we kept building on each other. Friend-1 came in steering the vibe right away, army not SWAT, here's the feel, dropping reference boards, and they basically handed me my whole client question list: the three main images they pictured, whether they wanted AWGE incorporated, realistic or glitchy, and most importantly what songs it was for. They kept hammering that the songs were the single most important thing to nail.

Friend-2 came through with strong references, and from there the direction came together fast between the three of us: army not SWAT, grunge, heavy haze and smoke, glitchy strobe, a hint of CCTV, soldiers with guns but only selectively and with red lasers off them, heavy helicopter visuals using the propeller blades to transition between scenes, even chopping the music video into black and white with boxes flashing over faces. They know this tour world and the vibe inside out, and they re-listened to the artist's discography to help tighten the aesthetic. From there I took the brainstorm and steered it into what we'd actually build.

Mood board

The board we pulled together, the references that locked the whole direction: red HUD scopes, surveillance and CCTV, black-and-white targeting, glitch, smoke. We built straight toward this.

References collected on Pinterest, full credit to the original creators. View the full board →

On the outreach the one who bit was LXND, within about five minutes on Instagram. And I didn't just text them back, I called them immediately, I wanted to actually talk it through. On the call they told me exactly what they wanted: soldiers walking toward the camera, a battlefield war-zone vibe, and to swag it out. This was for Thoto, an opener on the A$AP Rocky tour, so it was part of the Rocky tour, but the visuals were for Thoto's own set, their whole set start to finish, not Rocky's main show.

Right there on the call I came back with questions to pin it down, the environment, the colors, the tour colors, the soldiers' outfits, then followed up over text with the rest of Friend-1's list. I sent the mood board to confirm direction, they locked it to a single QR code, and on the songs they told me it was for every song of their set, so I realized this wasn't about scoring one track, they wanted a specific vibe carrying the whole set. They could tell I had it locked because they were looking right at the board. Friend-1 reminded me to nail the screen dimensions since those stage screens are huge, so I got the spec from LXND: 1920×1080, horizontal.

The proposal PDF

Then it was go time on the art. I started cooking thumbnails with one AI, but they weren't good enough, not realistic, with imperfections I didn't love. So I jumped to a better AI, ran the same prompts plus variations, and got results I loved. I'd take the one I loved and bring it to life as AI video with written prompts, directing it to do exactly what I wanted. That was my speed play, it kicked out usable motion fast so I could throw things together quickly, even if you could tell it was AI in spots.

First AI pass (scrapped, not real enough)

Better AI (the win)

The shot list

I couldn't just have them walking at the camera, I needed real angles.

Squad advancing toward camera (top-down)

Lone soldier walking toward camera into a red explosion

Black-ops face, quad night-vision goggles

Three soldiers through red smoke (top-down)

Advancing through the bomb blast

Black-ops face, second variation

Boot through the red smoke (side angle)

Tank tread, made but cut for time (saved for the next show)

The QR codes came through as a dark black-gray on transparent backgrounds, so to make sure they'd read clean against the dark scene, I brought one into vector software, vectorized it, and recolored it white. Then in image-editing software I built the thumbnail, carving a space into the back vest of one of the soldiers and placing that white QR code on it, and made the in and out frames, the start and end points, and exported them. Those frames went into a more powerful AI to animate. The trick is that if you only give the AI a single endpoint, the QR code gets mangled and won't scan, so I gave it both an in point and an out point. That more powerful AI holds the animation's integrity way better so the code stays sharp, the tradeoff being it takes longer to cook, which is exactly why I used the faster AI for the quick stuff and saved the powerful one for what had to stay crisp.

Before: the original dark gray, low contrast against the dark scene

Fixed: vectorized + recolored white in vector software

Image-editing software: the QR code built onto the back vest (in frame)

A more powerful AI: the out frame, animated in-point to out-point to keep the QR code scannable

Then I timed the QR code.

On top of that they wanted the QR code up once a minute, but my loop was only 25 seconds with the QR code at the very end, so in the final comp I duplicated the loop across the full ten minutes and stripped the QR code off the end of every other pass, so it surfaced about once a minute like they asked.

Comp 3: the final 10-minute assembly, the 25-second loop duplicated with the QR code kept on every other pass (the alternating colors) so it lands about once a minute, for exactly 10 minutes

It started all red, since we were going off the main A$AP Rocky vibe and my mood board was red. Then LXND and Casper both hit me to switch it to green, and I was already deep into building in red. My solve was two compositions in compositing software: the first was my movement and motion comp where I organized and chopped all the raw clips together, and the second was the final comp at 1920×1080, matched to the pixel map, where I dropped the first comp inside and applied a single hue rotate from red to green. Everything flipped in one move and I never had to redo a clip, I just kept building in red the whole time to match the board and flipped the switch at the end.

Comp 1: the raw clips clipped together in the compositing timeline

Then I hand-treated it.

In that same final comp I hand-treated the whole thing to sell it. I rode curves of calculated blur in and out at different points, ran a constant stream of monochrome grain that I kept pushing and pulling, varying the size and amount up and down so it breathed almost like CCTV footage, and gave the shots their own fades, coming up from black and falling back to black, so when the thing looped it never cut harshly, it just flowed.

Comp 2: the final comp, the movement comp dropped inside with the red-to-green hue rotate plus the grain and blur treatment stacked on top

The logo was its own little mission. Casper, on the media team, jumped in to help. They had the logo as a video on hand, and the source files (a GLB, SVG, or PNG) lived with another teammate, so they reached out to get them over, it just couldn't come through in time. The version I had was a 3D logo treatment with its own colors baked into it, which didn't match the look I was building, so I wanted to rebuild it from a clean PNG or SVG into a fresh 3D logo animation that fit the scene. I screenshotted a frame and used my own software to lift the logo out cleanly, converted it to a PNG, and built an SVG from it. The plan was a custom 3D version dropped into the scene with real-time environment lighting, color-matched to everything else so it'd sit perfectly in the shot. I had what I needed to build it, but I ran out of time, so the logo became a next-show thing, locked and ready.

Source: the LXND logo, sent as a 3D video with colors baked in

Pulled + cleaned to a PNG / SVG

16:9NextPlanned 3D logo asset matched with real-time environment lighting (for the next show).

Then we landed it. I talked to LXND, exported the final loop, then made a ten-minute version since that's how long it needed to run for Thoto's set. I asked them to send the Dropbox link, uploaded everything, named every file correctly and made sure they knew exactly which file was which, staying as organized as I could the whole way. Files were fully uploaded by 9:20 PM EST, so from the group chat kicking off at 5:44 PM EST to delivery, the entire thing was about a three-and-a-half-hour sprint. They loved it.

That obsession with staying organized is exactly why I built my agents the way I did. They take my raw files, look at the clips and understand what each one is, rename them properly, sort them into a full client file system, and stand up a shared delivery folder at the end, so even on a chaotic same-night turnaround nothing got lost and the client got something clean and labeled.

The client file system, every clip cleanly named and sorted

The Dropbox handoff, the final loop + 10-min version, cleanly named

The loop ran for Thoto's full set. Soldiers walking toward the camera, the full compositing treatment, the in-world QR code surfacing once a minute, all at 1920×1080, built and delivered in a single night.