The baile charme AI prompt

The baile charme AI trend looks simple, but getting a model to produce it takes a specific prompt: the right scene, a precise description of the step, and a way to keep your face. Below is the prompt structure our generator uses for the Viaduto Classic template, written out in English so you can use it in other tools.

The full prompt

Vertical handheld phone footage, night, a packed baile charme street party under a viaduct in Rio de Janeiro, strong floodlights behind the crowd. Camera movement, crowd, lighting and dance rhythm follow [reference video].

The lead dancer is the person in [photo 1]. Keep their face, hairstyle, skin tone and the clothes they are wearing.

The dancer looks back over their shoulder at the camera, then dances the charme step on the beat: shoulders bounce up and down, forearms swing left and right in front of the hips, knees soft and body leaning forward with the hands sliding past each other at waist height, then a small side step, ending with a turn. The crowd around them dances along.

Music and rhythm follow [reference audio]. No text, captions, stickers or watermarks on screen.

Replace [photo 1], [reference video] and [reference audio] with however your tool refers to attached files.

Why each part is there

1. The scene

"Vertical handheld phone footage, night, packed crowd, floodlights" is what makes the clip read as the trend. Leave out "handheld" and most models give you a smooth, cinematic camera that looks nothing like the original.

2. The lead dancer

Naming the photo as the lead dancer and asking to keep the face, hair, skin tone and clothes stops the model from blending you into the crowd or changing your outfit. For the Vovô no Baile template this line instead asks for you at 75 in a red shirt and suspenders.

3. The charme step

Models do not know what "charme" means, so the step is described in body parts and in order: look back, shoulders, forearms, forward lean with the hand slide, side step, turn. This mirrors the 12 second shot list of the original clip.

4. The sound and the clean frame

Linking the rhythm to the original audio keeps the moves on the beat. The last line removes the burned-in captions that many reference clips carry; without it, models often copy the text.

The 12 second shot list

• 0–2 s: look back at the camera.
• 2–6 s: shoulders bounce, forearms swing.
• 6–10 s: forward lean, hands slide past each other, crowd joins.
• 10–12 s: side step and turn.

Tips if you write your own

• Give the model a reference clip if it accepts one. Text alone rarely gets the step right.
• Avoid words like "edit", "replace" or "extend": some tools switch to editing mode and change the whole clip.
• Ask for 9:16 and 12 seconds; shorter clips cut the step before the turn.
• Do not use photos of celebrities or anyone who has not agreed.

Skip the prompt

The generator on this site already uses this prompt, a cleaned reference clip of the original shot and the viral sound. Upload a photo, pick a template and you are done.

Make my video