How I Turned Myself Into a 3-Person Flamenco Street Performance

How I Turned Myself Into a 3-Person Flamenco Street Performance

I filmed myself three times in my living room, once as each performer: a guitarist on the left, a flamenco dancer in the middle, and a second guitarist on the right. I merged the takes into one shot, then used AI to move all three of us onto a Spanish street at night and turn the middle version of me into a female flamenco dancer. Every strum and every move is my real performance. Here's how to do it yourself.

WHAT YOU NEED

  • A phone or camera on a tripod
  • A second camera or phone for close-ups (optional, but it makes the final video)
  • Your song, recorded ahead of time
  • CapCut for editing
  • A Higgsfield account for the AI image and the motion transfer

STEP 1: Record the song first
Record the full track before you film anything. You'll perform every part along to it, so all three takes line up with the same audio.

STEP 2: Plan your three positions
Before you hit record, decide where each performer stands: left, middle and right. Leave clear space between them so nobody overlaps. Picture the final scene. In mine, the left guitarist sits, the dancer is in the center, and the right guitarist stands with one foot up.

STEP 3: Film take one, the left guitarist
Lock the camera on a tripod, horizontal (16:9). It must not move at all for the rest of the shoot. Play the song and film yourself playing guitar in the left position.

STEP 4: Film take two, the dancer
Don't touch the camera or the lights. Move to the center and perform the dance to the same track. You don't need to be a trained flamenco dancer. Strong arm shapes, turns, sharp stops and clear poses are what the AI picks up best. Keep your whole body in frame, feet included.

STEP 5: Film take three, the right guitarist
Same camera and lighting. Move to the right side and play guitar again.

STEP 6: Shoot hand close-ups (optional, but worth it)
Film close-ups of your fretting hand and your strumming hand playing the real part. These go in the final split screen and prove the music is really you.

STEP 7: Merge the three takes into one shot
In CapCut, stack the three takes on separate tracks and sync them to the song using a clear beat or count-in. Mask each take so it only shows its own section of the frame: left, middle or right. Feather the mask edges so the seams disappear. Export it. You now have one video of three of you performing in the same room.

STEP 8: Grab the first frame
Move the playhead to 0:00 on the merged video and export that frame as a full-resolution image. This is your starting image for the AI.

STEP 9: Turn the still into the new scene in Higgsfield
Upload the still to Higgsfield and pick an image-editing model. I used GPT Image 2.5 at high quality, 2K, 16:9. Start the prompt by locking what has to stay the same, then describe the changes. Template:

"Edit this photo. Keep the exact same camera angle, framing and composition. Keep all three people in exactly the same positions and poses, with hands in the same places on the guitars. Turn the setting into a narrow Spanish old-town street at night, with warm street lamps, whitewashed walls, flower pots on wrought-iron balconies and wet cobblestones. Dress both guitarists in black flamenco shirts and trousers, playing acoustic flamenco guitars. Change the person in the middle into a female flamenco dancer in a red polka-dot ruffled flamenco dress, hair in a low bun, holding the exact same pose. Keep the dog in the same spot. Photorealistic, cinematic warm lighting, natural hands with five fingers."

STEP 10: Check the image before animating
Put it side by side with your original frame. Are all three people in the same spots? Are the guitar hands where yours were? Does the dancer's pose match your pose exactly? Fix one thing at a time and generate one image at a time. Don't burn credits rolling batches.

STEP 11: Transfer your motion onto the new image
In Higgsfield, open Genjutsu Motion Control. Add the AI image as the image reference and your merged three-person video as the video reference, and set it to 1080p. Prompt:

"Transfer the exact body, arm, hand and finger motion from the reference video onto all three people in the image. Keep their looks, outfits, the dress, the guitars and the street exactly as in the image. Keep the camera static and matching the image framing. Photorealistic."

STEP 12: Review and add your song back
Watch it all the way through. Check the strumming hands, the dancer's spins, and the dress moving naturally. Download the MP4, bring it into CapCut, and lay your original song over it.

STEP 13: Build the split screen and post
Make a 9:16 project and stack the clips: the reference video on top, the AI result in the middle, and your two hand close-ups side by side on the bottom. Label them "Reference," "Result" and "Performance." Export at 1080x1920 and post.

TIPS THAT SAVE CREDITS

  • The tripod is everything. If the camera moves even slightly between takes, the merge and the motion transfer both break.
  • Leave space between all three positions. Overlapping bodies are hard to mask and confuse the AI.
  • For the dancer, exaggerate your movements. Big, clear shapes transfer better than subtle ones.
  • Real things in your room can carry over. My dog wandered into the shot and made it into the flamenco street, which is one of the best parts.
  • Always start your image prompt by locking the camera and poses. Otherwise the AI redraws the whole scene from a new angle.
Back to blog