What the 58th Floor video actually is
“58th Floor” is a song by Anderson .Paak and Cordae from their joint album Heavy Is the Crown, released on 23 October 2026. COLORS filmed the two of them performing it in its studio and posted the performance on 30 October 2026: a seamless teal room, one microphone hanging between them, Anderson .Paak closing the hook and Cordae taking over for the verse.
The AI trend keeps that look and puts new people in it. Friends, couples, siblings and pet owners upload two photos and get a short clip of themselves trading the hook and the verse in a teal studio. The AI version is a new video of the people in the photos, not an edit of the real COLORS performance.
The 15 seconds, shot by shot
A 58th Floor AI clip has four shots with hard cuts, timed to the song. 0 to 5.5 seconds: a waist-up two-shot, photo 1 rapping the end of the hook to camera while photo 2 vibes beside them with their mouth closed, turning toward photo 1 at about 3.5 seconds as the verse is handed over. 5.5 to 8.7 seconds: a close-up of photo 2 taking the verse, eyes half closed, the mic at the edge of the frame.
8.7 to 11.5 seconds: a wide head-to-toe shot, photo 2 still rapping while photo 1 bounces and dances beside them. 11.5 to 15 seconds: a low two-shot, photo 2 in profile rapping straight at photo 1, who nods along with an impressed face. Only one mouth raps at a time. AI output varies from render to render, so watch the whole clip before you post.
Which photo goes where
Photo 2 does most of the rapping and gets the only close-up, so it needs the clearest face: sharp focus, even light, nothing over the eyes. Photo 1 is on screen the whole time but mostly reacting, which forgives a slightly softer picture.
Each person keeps their own face, hair and the outfit from their photo. If you want matching looks, pick photos where you are dressed the way you want to appear. The generator does not dress anyone in the artists’ caps, chains or sunglasses.
Common mistakes
Using one group selfie for both people: two faces in one image tend to blend, so crop each person into a picture of their own. Sunglasses, masks, low caps and heavy beauty filters: they hide the features the model needs to keep each face recognisable.
Forgetting the lyrics are explicit: the original sound stays on the clip, so choose Silent if you are posting somewhere that needs a clean version. Confusing it with other trends: the orange-booth Hotel Lobby rap video is a different trend with different songs and artists, and 58th Floor has nothing to do with a skyscraper or penthouse.
Describing the scene to a general AI video tool
If you build it yourself in a general image-to-video model such as Seedance, Kling or Dreamina, the words “58th Floor” on their own usually produce a high-rise apartment. Describe the scene instead: “a seamless teal-turquoise music studio, one vintage silver microphone hanging upside down on a long black cable between two rappers, soft shadowless light, vertical 9:16, not a penthouse or skyscraper.”
Then spell out who is where and when each mouth moves: person 1 left of the mic finishing a hook, person 2 right of the mic taking the verse, the listener’s mouth closed. Expect several tries. A dedicated two-photo generator does this staging for you and charges only for the render you keep.
Can I do it on my phone?
Yes. Open 58thfloorai.video in Safari or Chrome, choose the two photos from your camera roll, render, and save the MP4 to your phone. Then upload it to TikTok, Instagram or YouTube like any other video.
