Bulin 47's freestyle is one of October's biggest AI templates

A 19-second freestyle that Dominican artist Bulin 47 performed outside a restaurant in Madrid has become one of October's biggest AI video templates. MARCA reported on October 8 that the clip spread on TikTok and Instagram after content creator Kevin Wai walked up with a speaker playing a dembow beat and Bulin 47 freestyled back on the spot. Users then began using AI to replace his face while keeping his movements, gestures and voice.

The template works because almost nothing in it moves. Everything stays fixed except the performer: the street, the red lighting, the handheld camera, the crowd and the sound. A viewer recognizes the scene in a second no matter whose face is on it, which is what let the format travel.

What Is the Bulin 47 AI Video Template?

The original is a short, loud, hand-shot clip. Bulin 47, born Hancel Teodoro Vargas Reynoso, answered Kevin Wai's freestyle on the spot in early October, and the clip spread on TikTok and Instagram, according to invixtechnology. Early reports put the clip above 50 million views across platforms, and MARCA noted that other outlets said the original Instagram post passed 65 million views.

The hook became a single. "Dame Hombro," by Bulin 47, Kevin Wai and Breyco En Producidera, was released on streaming on October 7, 2026, per Qobuz, and a music video followed on the Bulin 47 YouTube channel.

Watch the official "Dame Hombro" video below. https://www.youtube.com/watch?v=VUIgT-E4IS0

The most-liked AI versions recast the same 19 seconds. Fan edits put footballers Lionel Messi, Cristiano Ronaldo, Neymar and Erling Haaland in Bulin 47's place, and versions with businesspeople, media figures and politicians followed, MARCA reported. Trendrecast, a vendor that sells a one-photo tool for this trend, says its own production logs show the clip rendered more than 500 times by October 12, 2026, with about 89% of renders succeeding (464 of 520).

There are three routes, with different results. A one-tap phone template, like the most-used CapCut version, drops a photo into a slot someone else built: it runs 10 seconds, about half the clip, and you get whatever the template maker made. Image-to-video, or reference generation, treats your photo as a hint and regenerates the scene; in the vendor's own test Kevin Wai mostly disappeared, the ceiling turned white and the framing changed. A video edit keeps the source frames and replaces only the performer, and it is the only route that returns the real scene with the original audio.

Edit mode holds the shot, but the vendor reports the face can drift toward the original performer, who wears sunglasses. Two details from the original performer carry over more often than expected: Trendrecast says the gold watch survived every prompt variant it tried, and the sunglasses still show up in some renders even when the photo has none. The template clip is 19 seconds, the length Trendrecast renders, and ByteDance's own Seedance 2.5 page separates reference-based generation from editing. A Seedance 2.5 prompting guide splits locked tasks, where the input drives the output, from unlocked reference tasks, and recommends keeping editing inputs under 20 seconds.

Some fans call the trend by the song's name, but the AI versions still use the street clip rather than the studio video, which has too many people in frame to work as a template. The same path looks repeatable: a street clip, a single, then a one-photo template within days, according to invixtechnology.

The trend is usable today. Trendrecast sells a 480p render of the full 19-second clip for $9.99, with renders taking about five to 20 minutes and no watermark. Its tool is independent and not affiliated with Bulin 47 or Kevin Wai, and neither are the other one-photo sites that have appeared around the trend.