The one-click way
Upload one photo on Burning Bridges AI, pay once (from $9.99), and you get the 15-second party clip with you in Drake's place — as a 9:16 vertical with the blurred-fill look of the viral posts, plus a 16:9 copy. It's usually ready in about 10 minutes and is emailed to you.
The manual way, step by step
- Pick the part of the video. Use the handheld party section of the official Burning Bridges video (from about 1:28). The swap clip everyone uses is the stretch around the cup raise and wave at about 2:58.
- Trim it to 15 seconds or less. Start a few seconds before the face cover and end on the walk-up close-up. Most character-swap models accept 10–15 s of reference video; keep the file under 10 MB.
- Get one clear photo. One person (or one character), face clearly visible, waist-up or full body. For a cartoon or character, a clean front view works best.
- Run a character-replace model. Give the model the clip and the photo and ask it to replace the main performer only. Wan 2.2 Animate (replace mode) did this with no prompt in our test; Higgsfield's Swap is what most viral versions used.
- Export vertical and post with the sound. Put the 16:9 result on a 9:16 canvas with a blurred copy behind it, then add the song in the app. In TikTok or Reels tap "Add sound", search "Burning Bridges Drake" and pick the version most posts use.
Which AI tools work (our tests, Oct 8, 2026)
We ran the same 15-second clip from the party section and the same AI-generated test photo through each route. One render per setup — no cherry-picking.
| Tool | Result | Notes |
|---|---|---|
| Wan 2.2 Animate (replace mode) | Worked, 2/2 | Kept the room, crowd, cup and every gesture; identity held through the close-up. No prompt. ~$0.94 per 15 s at 720p, 6–10 min. |
| Seedance 2.5 reference-to-video | Refused, 3/3 | “The output video may contain sensitive information” — the clip shows real, well-known people. Not billed. |
| Higgsfield Genjutsu (Swap) | Not tested by us | The tool behind the documented viral versions, per Top5Apps; supports mapping two characters. |
| Kling Motion Control | Not tested by us | Animates one character from your image with the clip's motion; with two or more people it follows the largest one (per Top5Apps). The room comes from your image, so the party itself is likely lost. |
The Burning Bridges AI prompt (as tested)
No one had published a prompt for this trend, so here is the exact one we ran with Seedance 2.5 (Video 1 = the clip, Image 1 = the photo). It is well-formed for reference-to-video models — but be warned: Seedance refused it every time because of who is in the source clip. Replace-mode models such as Wan 2.2 Animate skip the prompt entirely.
Replace the man in the black durag and black leather jacket holding a blue metal cup — the one who points at his friends, covers his face laughing, waves at the camera with the cup raised and finally walks up to the lens in Video 1 with the person in Image 1, matching his pose, gestures, cup and position exactly in every frame. The replacement keeps the face, hair, skin tone and build from Image 1 and wears the same clothes as in Image 1; keep the blue metal cup in his hand. Do NOT copy the original man's face, beard, durag or jacket. Video 1 is the exact footage to recreate. Copy it frame by frame: the same handheld camera path, sway, framing and timing, the same private dining room, chandeliers, green-tinted low light, crowd and every gesture, move for move: pointing, covering the face, the cup raise and the waves at the camera, walking up to the lens and turning away. One continuous take, no cuts. Everyone else in the room stays exactly as in Video 1 — same people, same clothes, same reactions; only the replaced person changes. Cinematic music-video footage, natural skin texture. No text, no captions, no logos, no watermarks.
Why each line is there
- Describe the performer by what he does (cup, face cover, wave, walk-up) so the model replaces the right person in a crowded frame.
- “Do NOT copy the original man's face” — reference videos pull the new person toward the original face.
- “One continuous take” — the scene has no cuts; models like to add them.
- “Everyone else stays exactly as in Video 1” — the joke only lands when only one person changes.
What photo works best
- One person in the photo, face clearly visible
- Waist-up or full body works best
- Pets and characters work — use a clear, front-facing image
- Only photos of yourself or people/characters you have permission to use
Common problems
- The cup melts into the hand. Usually a busy or low-resolution photo; use a cleaner one.
- The wrong person gets replaced. The model picks the biggest person in frame — keep your clip on the stretch where Drake is the main subject.
- The face drifts back toward the original. Use a photo with a larger, sharper face.
- A character looks too flat. Use a 3D-rendered image of it rather than a flat drawing, so it takes the room's light.
Label it, and mind the rights
Label the post as AI-generated and don't present it as a real scene from the video. The video belongs to OVO / Republic Records, and characters belong to their studios — use photos of yourself, people who agreed, or characters you have the right to use. Background on the trend: the Burning Bridges meme, explained. Sources: official video, Top5Apps, Wan 2.2.