58th Floor AI Video: two photos, one hanging mic, the COLORS rap duet, explained

A 58th Floor AI video is a 15-second vertical clip of two people rapping a duet in a bright teal studio, with a single silver microphone hanging on a long cable between them [official]. One person finishes the opening hook, the other turns to face them, takes the verse and gets the close-up, and the clip ends with the two of them facing each other, trading lines [official]. Nothing is filmed: two photos go in, and the template supplies the set, the mic, the shots, the moves and the song.

The clip behind it is Anderson .Paak and Cordae performing “58th Floor” on COLORS, released on 30 October 2026 as A COLORS SHOW [official]. The song is the opening track of their joint album Heavy Is the Crown, out 23 October 2026 [official], and COLORS framed the session around friendship: “Some people bring out your best. Others bring out your worst.” [official] On the set, .Paak stands on the left in a printed black T-shirt and silver chain, Cordae on the right in a black leather jacket — and that left/right casting is what the templates copy: photo 1 takes the left slot and the hook, photo 2 takes the right slot, the verse and the close-up [official].

The numbers behind it are modest, and they are dated reads, not a single count. On 3 November 2026 the COLORS show had 397,596 YouTube views and its vertical TikTok cut had 742,000 views [reported] — about 1.1 million combined. Two days later both had climbed: 542,939 on YouTube and 856,500 on the TikTok cut, read on 5 November [reported]. The hashtag was small too: #58thfloor held 89 videos and 72,600 views as read on 3 November [reported]. The trend is real but small, and it was still climbing when the templates arrived.

The template wave moved fast. The first two-photo AI templates of the performance appeared on 2 November [reported], Jump on the Trend went live with its version on 3 November — one of nine trends it launched that day, alongside 24K Magic, Stateside Dance, Bicep Heart and First Time in a Fancy Cafe [official] — and the trend tracker aivideotrend records the trend as started 3 November 2026, updated 5 November [official]. By 4 November at least seven template sites held a version: Jump on the Trend, SameVid, Starrd, aiMakeVideo, AI Photo Animator, Media.io and AI Effect Studio [official]. This page ships on day eight of the wave.

The trend also travels under the names “58th Floor COLORS trend”, “Anderson .Paak and Cordae AI trend”, “the teal studio trend” and #58thfloor [official]. Those are not five trends — they are five names for one performance, and this page covers all of them.

Every version that works hits the same beats: a waist-up two-shot, a turn, a close-up, a wide shot, and a face-to-face finish [official]. The timings below combine the two beat maps the template sites publish, which agree to within about a second [official]; the handover, not the tenths of a second, is what to check when a generation breaks.

0–4 seconds — the hook, both of you. Both of you waist-up at the hanging mic: the left person raps by the mic while the right person listens and moves to the beat [official]. This is the shot the free preview shows you, so check here that both faces look right before paying [official].

4–5 seconds — the turn and the handover. The two of you turn to face each other, and the right person takes over at about 5 seconds [official]. The handover is the joke the whole clip is built on — one rapper still riding the hook when the other turns and takes the verse, so suddenly it is a conversation between two friends [official]. If the turn lands late, the verse starts on the wrong face.

5–9 seconds — the verse close-up. Close-up of the right person rapping, with the mic at the edge of the frame [official]. The right-hand rapper fills the screen here, so their photo matters most: a sharp, front-facing face, or the close-up blows up a small cropped face and stops looking like them [official].

9–12 seconds — the wide shot. Wide shot, head to toe on the low platform: the right person raps and the left person dances [official]. This is the shot that needs a full-body photo — the wide frame shows everyone’s clothes and shoes, and anything the photo does not show gets continued in the same everyday style [official]. On the clip-remake route the outfits come from the session; on the generator routes they come from your photos [official].

12–15 seconds — face to face. Back to both from the waist up, or a low two-shot: the right person raps straight at the left person, who nods along, and the clip ends on the two of them facing each other [official]. SameVid says its workflow aligns every head bob, hand gesture and camera angle with the song’s 89 BPM beat [official].

The joke is the handover. Everyone has a friend they end up rapping with in the car; the clip just puts the two of you in the room where two professionals did it for real, and the turn at five seconds does the casting for you [official]. COLORS framed the session around friendship, and the template sites lean on exactly that: cast the friend who brings out your worst side [official].

The template has two parameters — who raps first, who takes the verse — and that pair is a variant machine. Any two adults fit the same teal scene, and the same mechanism as the tiny-desk and coldplay-kiss-cam templates applies: one fixed scene, swappable subjects [official]. The variants prove it: best friends and siblings trade the hook and the verse, roommates and brothers use it as an inside joke, hip-hop fans put themselves next to a friend in the booth, and the funniest versions pair opposites — a dad in a Hawaiian shirt next to his punk son, a coworker in a suit next to one in a hoodie [official]. One site even recasts the whole duet as a bride and groom, tux and lace dress, for engagement and anniversary posts [official].

The second accelerator is how little the format asks of you. No filming, no rapping, no lip-syncing: you upload photos and the template makes both of you perform the lines, mouth movements and gestures [official]. The solo fallback helps too — one photo plays both parts, like twins, on the generator routes, or takes the leather-jacket verse while the other performer stays as filmed on the clip-remake route [official].

The honest clock: the COLORS session went up on 30 October, the first templates appeared on 2 November, the wave went live on 3 November, and this page ships on day eight. That is early — but seven template sites already hold a version, and the source clip kept climbing while they arrived [official]. The durable part is not this song, it is the mechanism: one fixed studio, one hanging mic, two swappable friends, one handover.

There are two kinds of route, and they are not the same product. The clip-remake route puts your faces into the real COLORS footage; the generator routes rebuild the teal studio from scratch and put your photos in it. All of them make the same 15-second clip — the durations agree here, unlike some other meme trends — because they all copy the same session.

Route 1 — Jump on the Trend, the $6.99 clip remake (free preview). Pick “You and a friend” (or “Just you”), add one clear, front-facing photo per person, and in under a minute you get a free preview still of the two of you at the hanging mic; if it looks right, press Make my video and the 15-second video is 100 credits ($6.99), charged only when it is delivered [official]. No app, no prompt, no sign-up for the preview, and it works in a phone browser [official]. This route uses the real COLORS footage, not an AI lookalike of the studio: the outfits, the studio, the moves and the timing stay from the session, and only your faces and hair change [official]. Photo 1 goes on the person in the printed black T-shirt on the left, photo 2 on the person in the black leather jacket on the right, who gets the close-up [official]. With “Just you” your face and hair go on the one in the leather jacket and the other performer stays as filmed [official]. Photos are deleted after 30 days, never used for training, and a failed generation is refunded automatically [official]. The catch: no pets on this one — the trend is two people rapping with their hands at a mic, which does not work with a pet [official].

Route 2 — AI Photo Animator, the ~$3.33 generator route. Open the template — Wan 3.0 with the prompt already set up, 15 seconds, with sound — upload one photo for the left rapper and one for the right, and press Generate; their test video was ready in about 7 minutes and lands in My Videos, so you can close the tab [official]. One video takes 300 credits, about $3.33: new accounts get 75 free credits, and the one-time $5.99 trial pack (540 credits) covers one video and leaves 240 credits, with no subscription [official]. Download the MP4 without a watermark and post it [official]. Each person keeps their own clothes from their photo, down to the shoes [official] — this is the route to pick when the outfit contrast is the joke. Two separate photos are required: the template does not split a group photo, and a face cropped out of a group shot is usually too small for the close-up [official]. The exact cost shows on the Generate button, and failed renders are refunded automatically [official].

Route 3 — aiMakeVideo, the 15-or-30-seconds generator route. Add one or two photos, pick the length and the sound, then render, download and post — usually 10 to 30 minutes, 720p, 9:16, no watermark [official]. Fifteen seconds costs a flat 900 credits and 30 seconds costs 1,900; the smallest pack is $9.90 and covers one 15-second video, and a failed render returns to your balance [official]. New accounts get up to 10 free credits — nowhere near enough for a video [official]. One photo is enough: you rap with yourself, like twins; two photos make the duo [official]. Keep the original track, or choose Silent and add your own music when you post [official]. The 30-second version puts 15 more seconds before the 15-second part — the left person alone, a close-up of them, then a wide shot of both — but there is no 30-second example video yet [official].

Route 4 — Starrd, the 100–250-credit route with a 30-second cut. Pick the template and open its scene page, upload one or two photos (or a photo of your pet), and Starrd generates a 10–30-second video with Seedance 2.0 or Kling Motion Control, usually in a few minutes [official]. A video is 100–250 credits, no subscription, and credits never expire [official]. Starrd lists both the 15-second 58th Floor template and a separate “58th Floor Extended” template: the first 30 seconds at the hanging mic while your friend (or pet) vibes beside you, from the wide to the close-ups to the laugh [official]. Both templates carry a “Contains explicit language” flag [official]. No per-video dollar price could be verified for this route [unknown].

Route 5 — SameVid, the 400-credit template. SameVid’s 58th Floor AI template takes two portrait photos and makes the 15-second 9:16 duet: one character delivers the opening chorus hook before stepping back to hype up their partner, who steps into the spotlight to rap the verse with tight lip-sync [official]. The template is listed at 400 credits [official]. No dollar price for those credits could be verified [unknown].

Route 6 — AI Effect Studio, the bride-and-groom version. Same teal studio, same hanging mic, different joke: one of you in a black tux opens with a hand-waving verse, the other in a white lace wedding dress takes the close-up at the mic, then the camera pulls back to show you both on a round stage before you trade lines face to face [official]. Faces, hair and build come from your photos; the tux, the dress, the studio, the cuts and the sound come from the clip [official]. The 15-second clip costs 60 credits in 480p and 135 credits in 720p HD; pay for just this video from $9.99, once, and your first video is $4.99 [official]. Typical wait is 3–10 minutes and you can close the page [official]. Photos are deleted after 7 days, videos after 30 [official]. This version is made for two different people — using one person on both sides is not offered yet [official].

Route 7 — Media.io, the prompt route (price unverified). Media.io lists a 58th Floor trend — “Trade verses in a colorful hanging-microphone studio scene” — on its viral-trends gallery [official]. That is a prompt-and-reference workflow, not a one-tap clip remake: you bring the photos and the prompt, and the platform’s video models do the rest. No price for this route could be verified [unknown] — treat it as the do-it-yourself route, not the cheap route, until a price is confirmed.

The prompt route, by hand. No prompt exists from COLORS or the artists, and the template sites do not publish theirs [reported]. The condensed prompt below is our wording of the beat map the template sites publish, not a copy of any site’s text. Give it to any image-to-video model that takes two reference photos plus a reference video, with the COLORS session as the reference.

Recreate the reference video shot for shot: the same bright teal studio with one silver microphone hanging on a long cable between two people on a low platform, the same waist-up two-shots, close-ups and wide shots in the same order, the same fixed camera positions, the same single 15-second take structure, timing and moves. The ONLY change: the two rappers are now the two people in the uploaded photos.

The performers are those two people in EVERY frame: copy each exact face from their photo — face shape, eye color, eyebrows, nose, mouth, skin tone, age, hair and any beard, moustache, glasses or freckles exactly as in the photo. Never morph, never drift back to the faces of the men in the reference video. The clothes come from the photos: each person wears their own outfit, down to the shoes.

THE ACTION, matching the reference second by second: 0–4 seconds, both waist-up at the hanging mic, the left person raps the opening hook while the right person listens and moves to the beat. 4–5 seconds, they turn to face each other and the right person takes over at about 5 seconds. 5–9 seconds, close-up of the right person rapping at the mic. 9–12 seconds, wide shot head to toe on the low platform, the right person raps and the left person dances. 12–15 seconds, low two-shot, the right person raps straight at the left person, who nods along. Both keep rapping, nodding and gesturing on the beat the whole time, mouths moving on their own lines and resting on the other person's lines.

Plain studio performance: no audience, no instruments, no props beyond the hanging mic. Anatomy rules: one head each, two arms and two hands each, five fingers on each hand. Exactly two people in the studio. No dialogue beyond the song, no text, no captions, no stickers, no logos, no watermark.

For the 30-second extended version, add 15 seconds before the 15-second part: the left person alone at the mic, a close-up of them, then a wide shot of both — that is how the extended cut is structured on the routes that sell it [official]. For the bride-and-groom version, swap the outfits in the prompt: the left person in a black tux with a bow tie, the right person in a white lace wedding dress and veil, on a round teal stage [official]. Caption ideas from the template sites: tag your friend, say who brings out whose worst side, mention it is AI, and add #58thfloor and #heavyisthecrown [official].

Sound: the song is the session’s own recording of “58th Floor”, and the template videos keep it for their 15 seconds [official]. The song belongs to its rights holders, and the templates give you no rights to it — the fan-tool pages say so plainly, and they are not affiliated with Anderson .Paak, Cordae, the rights holders, COLORS, TikTok or Instagram [official]. If the lyrics are a problem where you post, use the Silent option and add your own music [official].

This is an early bet, not a proven search term. The session went up on 30 October and the template wave is eight days old. On November 11, 2026, Google autocomplete returned no suggestions for the “58th floor ai video” phrasing [reported]. The trend is real; the English demand for the AI-video phrasing is not yet established.

The demand signal is far weaker than the same batch’s Smiths word. The COLORS show had about 1.1 million combined views (397,596 on YouTube and 742,000 on its TikTok vertical cut, both read 3 November) [reported]. The I Love the Smiths scene from the same early-November batch had about 18 million views — Searchlight Pictures’ TikTok of the scene had 5.7 million views, and the best-known posts read add up to about 18.6 million (read 9 November) [reported]. Same batch, same tracker, roughly sixteen times the signal. Do not read the template count as demand: seven template sites does not mean seven times the audience.

The two view counts are dated reads, not a conflict. 397,596 YouTube views read on 3 November and 542,939 read on 5 November [reported] — the show gained about 145,000 views in two days. The TikTok vertical cut reads the same way: 742,000 on 3 November, 856,500 on 5 November [reported]. The 3 November reads are the ones this page leads with, because they are the reads from the day the wave went live.

The song has explicit lyrics — in the 15 seconds used. The template sites flag it themselves: Starrd marks both its 58th Floor templates “Contains explicit language”, and aiMakeVideo states plainly that the song has explicit language, including in the 15 seconds its page uses [official]. That rules out minors, school and work contexts, and playing it as-is in public. Use the Silent option where offered, or keep the post to an adult audience. Do not reprint the lyrics to work around it — the template sites do not, either [official].

The cheapest price is one route’s price, not the trend’s price. About $3.33 on AI Photo Animator, $6.99 on Jump on the Trend, a $9.90 smallest pack on aiMakeVideo, from $9.99 a video (first video $4.99) on AI Effect Studio [official] — and 100–250 credits on Starrd and 400 credits on SameVid with no verified dollar conversion [unknown]. Compare the finished-video price before you pay, not the credit number, and remember the free preview only exists on the clip-remake route: on the generator routes you pay before you see the result [official].

Pets work on some routes and not others. Starrd says its templates work with pets too [official]; Jump on the Trend says not on this one [official]; aiMakeVideo says not yet [official]. Check the route you picked before promising the dog a verse.

The original is a real performance, and the sound is not yours. The COLORS session is real footage of two real artists [official], and your version swaps faces onto it or rebuilds the set around your photos. Turn on the platform’s AI-generated label before posting, and never present it as the artists’ own clip [official]. Only use your own photo, or a friend’s with their yes; never a celebrity’s, a stranger’s, or a child’s photo — the template sites bar celebrities, public figures and children, and require 18+ [official].

The failure modes are face drift, lip-sync slip, and the close-up. The lips match roughly, not word for word — each person’s mouth moves on their turn and rests while the other raps, but the syllables will not line up exactly, especially on a fast verse [official]. The right-hand face is the one the close-up blows up, so a small or cropped face there is the most common break [official]. A group photo cut into two does not count as two photos [official]. In every case the fix the template sites give is the same: run it again for a new take [official].

Three prices could not be verified. Starrd’s per-video dollar price, SameVid’s dollar conversion for its 400 credits, and Media.io’s price for its 58th Floor template could not be found [unknown]. They are marked, not smoothed over.