Below is a full Talk Dirty dance AI prompt to paste into an image-to-video tool, a second-by-second version, what happened when we ran it once in Wan 3.0, and fixes to try. Rather not write a prompt? Make it here from one photo; the stage, the crew, the moves and the flames come from the trend's own clip.

  • For: image-to-video tools that take one photo
  • Length: written for a 10-second clip
  • Shape: set the tool to 9:16
From our Wan 3.0 prompt test: a woman with a red bob, glasses and a mustard sweater dances in front of a crew in yellow and black jackets and a pink TALK DIRTY screen
Copy and paste

The full Talk Dirty dance AI prompt

Paste this into an image-to-video tool together with one clear, front-facing photo, and set the shape to 9:16. Each part does a job; the notes below say which. In our run the model followed most of it, but not all (see below).

Our run came out 9:16 because we set the tool to 9:16. If your tool follows the photo's shape instead, start from a vertical photo.

Prompt

Vertical 9:16 handheld phone video filmed from the crowd at a live arena concert at night, one continuous 10-second shot with no cuts. The person from the photo is the lead dancer, wearing the same clothes as in the photo. 0-2 s: they walk to the front of the stage beside a crew of backup dancers in matching yellow and black jackets; a white spotlight drops on them from above and a big LED screen behind spells TALK DIRTY. 2-7 s: they dance in the centre of the stage in sync with the crew: hips swinging side to side, knees bent, arms rolling in front of the chest; tall orange flame jets shoot up from the edge of the stage. 7-10 s: the crew closes in behind them in a line and they keep dancing, confident and having fun. Keep the same face and hair as the photo the whole time. Realistic concert phone footage, stage lights and haze, five fingers on each hand, no captions or text other than the screen.

  1. The camera — "Handheld phone video filmed from the crowd" and "one continuous shot" give the look of a fan's concert clip without cuts.
  2. The timeline — The 0–2, 2–7 and 7–10 second parts set the order: walk-in and spotlight, the dance with the flames, then the crew closing in.
  3. The person — "Wearing the same clothes as in the photo" and "same face and hair" are there to keep the face and clothes from drifting.
  4. The clean-up line — Five fingers and no captions are there to cut down the most common AI video mistakes.
Second by second, then a real run

What the Talk Dirty dance prompt asks for, and what we got

The prompt describes three beats. We ran it once in Wan 3.0 with a portrait photo to see which of them a video model actually follows.

The 10 seconds, beat by beat

If your tool takes shorter clips, keep the order and squeeze the timing.

  1. 0 s–2 s

    Walk to the front of the stage with the crew; the spotlight drops and the screen reads TALK DIRTY.

  2. 2 s–7 s

    Dance in the middle in sync with the crew while flame jets shoot up.

  3. 7 s–10 s

    The crew lines up behind you and the dance keeps going.

Our run in Wan 3.0

The face, the red bob, the glasses and the mustard sweater stayed right, she walked into a white spotlight, a pink TALK DIRTY screen lit up behind her and flame jets fired behind the stage. But the clip opened on a wide shot with her photo shown on a side screen, the crew came out as all women, and the dance was generic hip and arm moves, not the trend's routine.

Prompt we used

Vertical 9:16 handheld phone video filmed from the crowd at a live arena concert at night, one continuous 10-second shot with no cuts. The person from the photo is the lead dancer, wearing the same clothes as in the photo. 0-2 s: they walk to the front of the stage beside a crew of backup dancers in matching yellow and black jackets; a white spotlight drops on them from above and a big LED screen behind spells TALK DIRTY. 2-7 s: they dance in the centre of the stage in sync with the crew: hips swinging side to side, knees bent, arms rolling in front of the chest; tall orange flame jets shoot up from the edge of the stage. 7-10 s: the crew closes in behind them in a line and they keep dancing, confident and having fun. Keep the same face and hair as the photo the whole time. Realistic concert phone footage, stage lights and haze, five fingers on each hand, no captions or text other than the screen.

Settings
Image to video, 480P, 9:16, 10 seconds, one portrait photo, first result
Wait
About 4.5 minutes
Came out right
Face, hair, glasses and sweater; the spotlight walk-in; the TALK DIRTY screen; a crew in yellow and black jackets; flame jets from about 4 seconds
Missed
Her photo on a side screen in the first second; black shorts and boots she wasn't wearing; an all-women crew; flames behind the stage, not at its edge; generic moves instead of the routine
A woman with a red bob, glasses and a mustard sweater dances between crew members in yellow and black jackets, flame jets behind them, from the Wan 3.0 test run
The portrait used for the test: a woman with a red bob and glasses in a mustard sweater against a grey wall
Photo we used

Fixing common problems with the Talk Dirty dance prompt

We have not tested these fixes. Try one change per run, so you know what helped.

  • Your photo shows up on a screen or as a poster

    Try next
    Add "the photo is only a reference for the person; do not show it in the video"
    Leave alone
    The timeline
  • The dance is generic, not the trend's moves

    Try next
    Describe the moves you want in more detail, beat by beat; a prompt can't copy a routine it hasn't seen, so expect it to stay close rather than exact
    Leave alone
    The camera line
  • The crew looks different from the trend

    Try next
    Spell out the crew: "a mixed group of men and women in matching yellow and black jackets and black trousers"
    Leave alone
    Your photo
  • The flames are in the wrong place

    Try next
    Say where: "flame jets shoot up from the front edge of the stage, between the dancers and the crowd"
    Leave alone
    The spotlight line
  • Clothes you weren't wearing appear

    Try next
    Use a full-body photo, or name the whole outfit in the prompt
    Leave alone
    "Same face and hair as the photo"
  • The screen text is garbled

    Try next
    Keep TALK DIRTY in capitals in quotes, or drop the screen and say "a big LED screen with stage visuals"
    Leave alone
    The rest of the stage

The Talk Dirty dance prompt vs one photo here

A prompt lets you describe any stage and crew, and the dance is the model's own. One photo here gives you the trend's own clip and routine, without writing anything.

This promptIn an image-to-video toolOne photo hereNo promptUse my photo
What you writeThe prompt above, adjusted until it looks rightNothing
The danceWhatever the model comes up with; in our run, generic movesThe routine from the trend's clip
The stage and crewWhat the model imagines from your wordsThe trend's stage, crew and flame jets
Your photoThe person, and sometimes shown on a screen, as in our runGives the face, hair and top; the stage comes from the clip
ShapeWhatever you set; ours was 9:16 at 480P10 seconds, 9:16, 720p

Make the Talk Dirty dance from one photo instead

Add one photo for the lead dancer, and the stage, crew, moves and flames come from the trend's clip.

Example · 10 s · 9:16

One person per photo, face clearly visible. JPG, PNG or WebP up to 10 MB.

  • No app
  • No watermark
  • Credits back if a video fails
See pricing

Talk Dirty dance AI prompt questions

Rather not write a prompt?

One photo, the trend's own stage and routine.