Quick answer
Starrd's 58th Floor template turns two photos into Anderson .Paak and Cordae's viral '58th Floor' COLORS session. Photo 1 stands on the left and finishes the hook, photo 2 stands on the right and takes the verse, rapping straight at photo 1, in the seamless teal studio with the vintage microphone hanging between them. All four shots are cut where the original cuts, both of you lip-sync the original audio on the beat, and you each keep your own face, hair and outfit. 15 seconds, vertical 9:16, Seedance 2.5, 150 credits (about $7.50). The 58th Floor Extended template is the first 30 seconds of the session: photo 1 raps the whole verse and hook while photo 2 vibes beside them, 250 credits (about $12.50).
Two photos in, and out comes you and a friend in Anderson .Paak and Cordae's "58th Floor" COLORS session: the teal studio, the mic hanging between you, one of you finishing the hook and the other taking the verse. Every line lip-synced, every cut on the beat. Here it is:
58th Floor
Two photos. One of you finishes the hook, the other takes the verse, in the teal COLORS studio. 15 seconds, lip-synced.
What You Get
- The handoff. The 15 seconds where the mic changes hands: photo 1 finishes the hook, the two of you turn to face each other, and photo 2 takes the verse.
- Four shots, cut on the beat. A waist-up two-shot, a close-up of photo 2 with the mic at the edge of the frame, a wide from head to toe, and a low two-shot of photo 2 rapping straight at photo 1. Every cut lands where the original cuts.
- Real lip sync. Each of you mouths your own part of the original audio, and whoever isn't rapping keeps their mouth closed and vibes.
- Your own outfits. Faces, hair and the clothes from your photos, down to the shoes.
- The COLORS look. A seamless teal studio and one vintage microphone hanging between you.
- 15 seconds, vertical 9:16, ready for TikTok, Reels and Shorts. Want more? The 30-second extended cut is further down.
What the Video Actually Is
COLORS posted the clip on October 30, 2026 from @colorsxstudios, captioned "Some people bring out your best. Others bring out your worst." It passed 650,000 plays, 110,000 likes and nearly 12,000 saves in three days. It's Anderson .Paak and Cordae performing "58th Floor", from their joint album Heavy Is the Crown, in COLORS' signature setup: a single-colour studio with nothing in it but the artists and one microphone hanging from the ceiling.
The session is a friendship in 54 seconds. .Paak opens with the verse and the hook ("Murderation on the beat, highly favored energy… they don't make 'em like us, most certainly") while Cordae stands beside him nodding along. Then they turn to face each other, the mic changes hands, and Cordae raps his verse straight at .Paak, who grins and nods with a scrunched-up "that's hard" face.
Why it suits AI: the set is the whole look and almost nobody has one, and the joke works best with two friends. Two photos put you both at the mic without memorising a verse.
The 15 Seconds, Beat by Beat
The song runs at about 89 BPM, so every line takes one bar, about 2.7 seconds.
| Time | Shot | What happens |
|---|---|---|
| 0:00 | Waist-up two-shot | Photo 1 raps the end of the hook to the camera; photo 2 vibes, mouth closed |
| 3.5s | Same shot | Photo 2 turns to face photo 1 as the verse is handed over |
| 4.9s | Same shot | Photo 2 starts rapping |
| 5.5s | Close-up of photo 2, mic at the left edge | Photo 2 takes the verse, eyes half closed, a hand up into the frame |
| 8.7s | Wide, head to toe | Photo 2 raps, photo 1 dances beside them |
| 11.5s | Low two-shot | Photo 2 raps straight at photo 1, who nods along, impressed |
Who plays who: photo 1 finishes the hook, photo 2 raps the verse. Put the friend who wants the spotlight second.
The 30-Second Extended Cut
The extended cut is the first 30 seconds of the session, the part where .Paak carries it. Photo 1 raps the whole opening verse and the hook across five shots: the wide, a medium shot and a close-up of photo 1 alone, the wide again, and the waist-up two-shot. Photo 2 is the hype man beside them, nodding along with their mouth closed, until 21 seconds, when photo 1 swings an arm high and photo 2 throws a hand to their forehead with an open-mouthed laugh.
58th Floor Extended
The first 30 seconds of the session: you rap the verse and hook at the hanging mic while your friend vibes beside you. Lip-synced.
Your Look, Their Studio
- Your own face, hair and outfit. Nobody gets .Paak's cap and sunglasses, his chain or Cordae's leather jacket.
- Your own shoes. The wide shots show you head to toe, so the full outfit from your photo comes along. If your photo stops at the waist, the rest is filled in to match.
- Their studio. The teal backdrop and the hanging microphone, in every shot, right where they are in the original.
Full-length photos where you're standing and facing the camera work best, because the wide shots show your whole outfit. Clear, well-lit faces matter most for the close-ups. If you and your friend are dressed alike, that's fine: the template tells you apart by hairstyle as well as clothes.
The Fastest Way
58th Floor
Two photos, the 15-second handoff with the original audio, both of you lip-synced.
Two photos, 150 credits (about $7.50) for the 15-second handoff or 250 credits (about $12.50) for the 30-second extended cut, no prompt. Both render on Seedance 2.5 at 480p, vertical 9:16.
Build It Yourself
1. Turn the clip into a depth map. Run the session through a depth-estimation model (Depth Anything V2 works) and set the contrast per shot, because the wides and the close-ups have very different depth ranges. A depth map carries the moves, the camera and the cut points, but no faces, clothes or colour, so the original artists can't leak into your video.
2. Scrub the shapes that would leak. Even in a depth map, .Paak's cap and Cordae's tall afro come through as outlines, and your cast inherits them. Use a segmentation model (CLIPSeg works) to find hair, hats, sunglasses and chains in every frame, then smooth the heads down to plain, slightly smaller domes and blur the glasses and the chain away. Scale the smoothing to the shot: a filter strong enough for a close-up erases the whole head in a wide.
3. Make one storyboard sheet with a panel per camera setup. The session boils down to six setups: the wide, the waist-up two-shot, photo 1's medium shot and close-up, photo 2's close-up and the low two-shot. Generate them all from the two photos in one teal studio, with a depth-map sheet of the original framings as a layout guide so each panel matches its shot.
A close-up of the person from reference image 2 from the chest up in a seamless, evenly lit teal-turquoise studio, a vintage silver large-diaphragm condenser microphone hanging upside down on a long black cable at the LEFT edge of the frame, seen in three-quarter view with the head turned toward the left edge of the frame, eyes half closed, rapping, mouth open mid-word, wearing exactly the clothes from their photo. Soft, shadowless studio light, a touch of grain. No text.
4. Drive it with the depth map and the song. Put the original audio inside the depth video so the model hears the rap it has to match. Feed the sheet, both photos and the depth video to a model that copies motion from a reference video (we recommend Seedance 2.5). Then write three things it can't get from a depth map:
- A timed shot list: the framing for each shot and who is in it.
- The rap windows: exactly when each person's mouth moves, and that it stays closed otherwise.
- The faces: the rapper's confidence, the listener's nods and reactions.
Render it silent and lay the full original audio back on at the end.
Common Mistakes
Picking the first 15 seconds for a two-person video. .Paak raps the whole first 31 seconds, so photo 2 would never open their mouth. The handoff is the 15 seconds where you both rap.
Driving with the raw clip. The cap, the sunglasses, the chain, the leather jacket and the original faces come straight through. Use the depth map.
Leaving the cap in the depth map. Its brim reads as a hat, and your rapper gets one they never wore.
Letting both mouths move. Without rap windows, the listener mouths along to the other person's lines. Say who raps when, and that the other mouth stays closed.
Moving the microphone. It hangs dead centre between the two of you in the wides, at the edge of the frame in the close-ups. If the sheet puts it anywhere else, the video does too.
FAQ
How many photos? Exactly two: photo 1 on the left, photo 2 on the right.
Which one should I pick? 58th Floor if you both want to rap; 58th Floor Extended if one of you wants the whole verse.
How long is it? 15 seconds, or 30 seconds for the extended cut, vertical 9:16.
What does it cost? 150 credits (about $7.50), or 250 credits (about $12.50) for the extended cut.
Related Reading
- How to Make the Hotel Lobby AI Video (A COLORS SHOW): the other COLORS session, Quavo and Takeoff's "Hotel Lobby", starring you and a friend
- How to Make the Clayface AI Video: rap Drake's "Solar Eclipse" to the camera in a nightclub VIP booth
- How to Make the Outstanding Tiny Desk AI Video: your own Tiny Desk band, shot for shot
- How to Make the Raindance AI Video: Dave and Tems' "Raindance" as your own music video for two
- Viral AI Video Trends 2026: every trend we've turned into a template this year
