A working 58th Floor AI prompt describes the scene instead of naming it: a seamless teal studio, one vintage microphone hanging between two rappers, photo 1 on the left finishing the hook, photo 2 on the right taking the verse, one mouth moving at a time, and “not a penthouse or skyscraper”. Copy the prompts below.
One clear adult or pet per photo, face visible. Separate selfies are fine. JPG, PNG or WebP, 10 MB max, 300 × 300 pixels minimum.
The original “58th Floor” sound is kept by default; it contains explicit language. Pick Silent to export the clip with no soundtrack.
10 credits · Seedance 2.0 Fast · 15s · vertical MP4 · See pricing
Written for image-to-video models that take reference photos (Seedance 2.x, Dreamina, Kling). Replace the bracketed parts.
A photorealistic vertical 9:16 live studio rap performance, about 15 seconds, four shots with hard cuts on the beat. A seamless, evenly lit teal-turquoise cyclorama studio; one vintage silver condenser microphone hangs upside down on a long black cable between the two performers at mouth height; nothing else in the room. Rapper 1 is the person in image 1 [describe hair and outfit] and always stands on the LEFT of the mic. Rapper 2 is the person in image 2 [describe hair and outfit] and always stands on the RIGHT. 0–5.5 s waist-up two-shot: Rapper 1 raps the end of the hook to camera, Rapper 2 vibes with mouth closed and turns toward Rapper 1 at 3.5 s. 5.5–8.7 s close-up of Rapper 2 taking the verse, eyes half closed, the mic at the left edge of the frame. 8.7–11.5 s wide, head to toe: Rapper 2 raps while Rapper 1 bounces beside them, mouth closed. 11.5–15 s low waist-up two-shot: Rapper 2 in profile raps at Rapper 1, who nods with an impressed face, mouth closed. Only one mouth moves at a time. Keep each person’s exact face, hair and clothes from their photo. Soft shadowless light, a touch of grain. This is a music studio, not a penthouse or skyscraper.
Most models will not hit these timings exactly without a motion reference; expect a few tries.
A photorealistic vertical 9:16 storyboard sheet of six film stills in a 2×3 grid, thin black lines between panels, no text. All six are from one live performance in the same seamless teal-turquoise studio with one vintage silver microphone hanging between two rappers. Rapper 1 (image 1) always on the left, Rapper 2 (image 2) always on the right, each in exactly the clothes from their photo. Panels: a wide head-to-toe shot; a waist-up two-shot; Rapper 1 alone waist-up; Rapper 1 close-up; Rapper 2 close-up with the mic at the left edge; a low two-shot with Rapper 2 in profile rapping at Rapper 1.
Feed the sheet back in as a reference image for the video so the faces and room stay consistent.
Avoid: skyscraper, penthouse, city view, windows, orange booth, extra people, text, captions, logos, watermarks, both rappers mouthing at once, mixed-up faces or clothes, extra limbs, morphing.
Add this to the negative prompt or the end of the main prompt.
“58th Floor” is a song title, not a description. Image and video models read it literally and draw a high-rise apartment, a city view or a luxury penthouse. Describe what is in the frame instead: the teal room, the hanging mic, the two people and who raps when.
If a chatbot writes the prompt for you, tell it that 58th Floor is the Anderson .Paak and Cordae COLORS performance, a teal studio with one hanging mic, not a building.
The template sites do not rely on a prompt alone. The common recipe is a two-step render: an image model first draws a storyboard of the two people in the teal studio, then a video model renders from that storyboard, the two photos and a motion reference clip that carries the song’s timing, so each mouth opens on its own lines. The original sound is added to the finished file.
That motion reference is the hard part to reproduce at home, which is why the hand-over and the lip sync are where DIY attempts usually drift.
How to make it, which photo goes where, pets, the song, the prompt, prices and every other 58th Floor AI tool.
Two photos in, a 15-second teal-studio duo out, made in your browser.
Four steps, the 15-second shot list and the common mistakes.
Starrd, MakeViral, VideoSi, Flowlio, AI Meme Videos and us, side by side.
Anderson .Paak and Cordae’s opening track, the credits and the COLORS session.
What ChatGPT can do for the trend, and what renders the clip.
Add two photos, check the price on the button and get your 15-second 58th Floor AI clip in a few minutes.
