How to Write Jazz Dance AI Video Prompts? NadouPro Canvas Tutorial: Scenes, Camera Angles, Costumes, Motion

Key Takeaways

• Jazz dance AI videos commonly have six types of problems: distorted limbs, movement that misses the beat, duets with no interaction, face and outfit changes after camera cuts, unrealistic costume motion, and monotonous camera movement. All of them can be addressed at the prompt and workflow level.

• Motion descriptions should be broken down to the body-part level: body part + action + force and speed + costume interaction. Vague wording such as “dance jazz” is an ineffective prompt.

• The most stable path for duet dance is Motion Control: feed in a real duet dance reference video so the model can extract facial expressions and body movements and apply them to the characters.

• More than 20 ready-to-adapt prompt templates for scenes, movement, camera work, and costumes.

• Case demo: generate a 30-second jazz duet on the NadouPro canvas, creating three 10-second clips separately and then stitching them together in Clip.

1. Common Problems in Jazz Dance AI Videos

Jazz dance has fast rhythms, wide-ranging movements, and many costume and styling elements, making it a relatively failure-prone subject in AI video. The common problems fall into six categories:

Distorted limbs. The number of fingers is wrong, joints bend in strange directions, and arm trajectories twist during fast hand swings. Solution direction: break motion descriptions down to body parts, choose a model with strong dynamic performance, and use Motion Control with a real reference video.

Stiff movement that misses the beat. The movement feels like it is being pulled by strings, without the sharp stops and accents unique to jazz dance. Solution direction: clearly describe speed and pauses in the motion prompt, and generate with an audio-visual synchronization model.

Duet dancers each doing their own thing. Two dancers move in parallel with no interaction or eye contact, and may even pass through each other’s bodies. Solution direction: clearly state the interaction relationship in the prompt, and use Motion Control mode for duet movement.

Face and outfit change after a camera cut. The dancer’s face, hairstyle, and costume drift between different shots. Solution direction: create finalized look assets first, reference the same asset set for every shot in the full video, and use a swatch to constrain the color tone.

Unrealistic costume motion. Fringe breaks apart, sequins flicker unpredictably, and skirt movement does not match the rotation direction. Solution direction: specify materials in the costume prompt, and include the costume’s linked response in the motion prompt.

Monotonous camera movement. The whole video is a flat wide view with no change in shot size. Solution direction: design a shot list first and generate shot by shot according to shot size.

Below, we will go through solutions and ready-to-use templates from five angles: prompts, scenes, motion, camera work, and costumes.

2. Overall Prompt Structure: Build the Right Framework First

No matter which step you are working on, write each prompt with this structure:

Subject (character and costume) + motion (body parts and force) + scene environment + lighting and color tone + camera (shot size and movement) + style

Here is a complete jazz dance shot prompt for comparison:

Two dancers in black sequined jazz costumes perform on a theater stage, shoulder shimmies followed by synchronized hip circles, the fringe skirt spreads with the rotation, red curtain and side backlight, medium shot, the camera slowly orbits, cinematic quality

Three common mistakes: only writing adjectives such as “sexy,” “handsome,” or “energetic”; piling up movement words without describing the characters and costumes; cramming more than two conflicting camera movements into one prompt.

3. Scene Prompts: Five Common Jazz Dance Settings

The scene determines the lighting and color style. Here are five types of scenes, with two templates each. You can use them directly by changing the details.

Theater stage.

Theater stage, a single spotlight shines down from directly overhead, thin fog on the floor, red velvet curtain, black mirrored floor reflecting human silhouettes, dramatic side light

Grand theater stage, warm golden lighting array, curtain opened, thin fog drifting through the beams, strong stage depth, backlit silhouette effect

Rehearsal room.

Dance rehearsal room, a full wall of floor-to-ceiling mirrors, afternoon sunlight slanting into the room, reflective wooden floor, barre in the corner, natural-light texture

Rehearsal room at night, only one row of ceiling lights is on, mirrors reflect a repeated space, cool color tone, quiet atmosphere

Street and underpass settings (street jazz).

Urban underpass, concrete walls with graffiti, side backlight at dusk, distant pedestrians blurred, handheld camera texture

Street corner at night, one streetlamp overhead, wet ground reflecting light, colors from neon signs reflected on the floor

Retro jazz bar.

Small stage in a retro jazz bar, warm neon sign, smoke slowly rising, musician silhouettes in the background, amber color tone

Old-style cabaret stage, golden curtain fringe, circular follow spot, warm tungsten light, nostalgic film feeling

Concert-level main stage.

Large concert stage, giant screen in the background, an array of beam lights sweeping across, smoke filling the air, reflective stage floor, contrast between cool and warm light

Outdoor music festival main stage, night background, stage lighting array projecting columns of light into the sky, audience light spots blurred

After the scene is set, use the Image node’s Relight mode to make a second adjustment to the lighting feel, and use a swatch to save the colors of this scene so later shots can align with it.

4. Motion Description Prompts: The Most Important Part of the Whole Article

Write motion prompts using “body part + specific action + force and speed + direction + costume or environment interaction.” “Danced a jazz routine” is an ineffective prompt. “The dancer quickly whips their head, the shoulders shimmy continuously, and the movement pauses sharply on the beat” is effective wording.

Solo motion templates.

The dancer quickly shimmies the shoulders, whips the head to the side, long hair flying with the motion, one hand holding the brim of the hat, the movement stops abruptly on the beat, medium shot

The dancer walks forward with jazz steps, hips swinging with every step, shoulders alternating forward and back, eyes looking straight into the camera, full shot, camera slowly dollies backward

The dancer spins twice in place and then freezes abruptly, the fringe skirt spreading into a fan shape, chest rising and falling with the breath, low-angle shot

The dancer’s body rises and falls like a wave from feet to head, arms extending upward along the body, the movement remains continuous, medium close-up

The dancer kicks to the side, the leg rising above hip height, upper body leaning in the opposite direction to maintain balance, the movement crisp and powerful, full shot from a side camera position

Duet motion templates.

Two dancers face each other, performing mirrored synchronized shoulder shimmies and hip circles, their gaze locked on each other the entire time, medium shot

Two dancers move side by side, the movement completely synchronized, whipping their heads and spinning at the same time, fringe skirts spreading together, full shot

The female dancer walks around the male dancer, while the male dancer stays in the center and only turns his upper body, with a clear lead-and-follow relationship, the camera orbits around both of them

Two dancers walk toward each other, crossing at the closest distance, shoulders brushing past each other, the hems of their clothes lifting airflow, the camera tracks sideways to follow

Two dancers stand back to back, slowly sliding apart to both sides along each other’s backs, then finally turn back at the same time to make eye contact, close-up cutting to full shot

One reminder: difficult body-contact moves such as lifts and tosses are currently prone to deformation in AI generation. For duet dance, it is better to keep interaction at the level of blocking, mirroring, leading and following, and brushing past each other.

If you want fully professional movement, use Motion Control. Upload a real dance video or character image, and Motion Control mode will extract the facial expressions and body movements and apply them to the target character. If you have a choreography demo video, use it. This is currently the most stable generation method for dance subjects.

5. Camera Prompts: Let the Shot Size Follow the Movement

The fixed rule for dance shots is: use full shots to show formation and movement range, medium shots to show upper-body rhythm and duet interaction, and close-ups to show facial expressions, hands, and shoulder details.

Full shot, low-angle shot, two dancers standing at center stage, spotlight following the movement, background darkened

Medium shot, the camera slowly orbits around two dancers, capturing their face-to-face interaction, shallow depth of field blurring the background

Close-up, the dancer’s shoulders shimmy quickly, hand holding the brim of the hat, strong sense of rhythm

High-angle full shot from above, two dancers standing in a diagonal arrangement, clear reflection on the mirrored floor, camera slowly rotating

Handheld-texture medium shot, the camera sways slightly with the dance steps, documentary style

If you are unsure about the camera plan, use the Multi-Camera 9-Grid in Multi-Grid to generate nine camera-angle previews at once. For shots with complex multi-character blocking, use 3D Director to lay out the character positions and camera movement first. It supports pose presets and skeletal motion adjustment, and its multi-camera timeline can output up to 90 seconds, so camera movement and blocking can both be locked in advance.

6. Costume Prompts: Describe Materials and Dynamic Interaction

Jazz dance costumes are a visual focus in themselves. A costume prompt should describe three things: style and silhouette, material reflectivity, and how it behaves in motion.

Black sequined one-piece jazz costume, sequins reflecting finely under the light, mesh long sleeves, fringe hem, black jazz shoes

Gold fringe skirt, fringe reaching below the knees, spreading into an arc when spinning, warm golden reflection

Black suit, white shirt, wide-brim hat, satin tie, retro jazz style, crisp fabric

Black practice leotard with high-waisted fitted pants, minimalist practice outfit, matte close-fitting fabric

Duet costume pairing: male dancer in a black suit and wide-brim hat, female dancer in a red fringe skirt, black-and-red contrast, unified style

Costume consistency is solved through assets: first use the Image node to create costume setting images, then use Image Fusion mode to combine the character image and costume image into finalized look images, and reference the same finalized asset set throughout the full video. During the finalized-look stage, you can enable photo parameters to fine-tune the texture. Note that this feature is only available in Text to Image and Image Fusion modes when selecting the Nano Banana 2, Nano Banana Pro, or GPT Image 2 models.

7. Case Study: Generate a 30-Second Jazz Duet on the NadouPro Canvas

Using a typical scene as an example, here is the full production workflow for “night theater stage, black-and-red duet costumes, jazz duet.” Text to Video generates 1- to 15-second clips at a time, so a 30-second finished video is made by generating three 10-second clips separately and stitching them together at the end.

Step 1: Create duet assets. Use the Image node to generate three-view character sheets for the male and female dancers respectively, create two costume setting images for the black suit and red fringe skirt, use Image Fusion mode to composite finalized look images for both dancers, and save the swatch.

Step 2: Define three shots.

Shot 1 (0–10 seconds): full shot, two dancers walk from deep stage toward the spotlight, stop in position, then perform synchronized shoulder shimmies while facing each other

Shot 2 (10–20 seconds): medium orbit shot, the two dancers perform mirrored synchronized hip circles and spins, the fringe skirt spreads out

Shot 3 (20–30 seconds): low-angle shot, the two dancers whip their heads at the same time and then freeze abruptly, the spotlight narrows into one beam

Step 3: Feed in motion reference. Give a real duet jazz dance reference video to Motion Control mode, extracting facial expressions and body movements and applying them to the two dancers. The three shots correspond to three sections of the reference video.

Step 4: Generate the clips. In Multi-reference mode, reference the two finalized look images and the stage scene image at the same time. First use Seedance 2.0 mini to quickly preview the motion direction. After confirming it, switch to Happy Horse 1.1, which has stronger dynamic performance, for final generation, 10 seconds per section. All video models support audio-visual synchronization, so footsteps and clothing friction sounds will be generated together with the visuals.

Step 5: Add music. Use the Audio node’s Text to Music to generate a 30-second jazz dance track, clearly writing “brass-led, dense drum beats, with improvised piano, fast tempo.”

Step 6: Stitch the finished video. Connect the three video clips and the music into Clip, stitch them online, adjust the rhythm points, then export after improving quality with Frame Interpolation / Ultra HD.

FAQ

Q: The two dancers always pass through each other or each dance separately. What should I do?
The most stable method is to feed a duet reference dance video into Motion Control mode. For pure text generation, clearly write interaction relationships such as “face to face,” “mirrored synchronization,” and “moving around each other” into the prompt. Avoid difficult contact moves such as lifts.

Q: How do I keep both people’s faces consistent over 30 seconds?
Create one set of finalized look assets for each person, use Multi-reference mode to reference the same asset set in every shot, and use a swatch to constrain the overall color tone.

Q: Can a single shot be longer than 15 seconds?
Text to Video generates 1 to 15 seconds at a time. To extend it, use video editing mode to extend the duration of an existing clip, or generate multiple clips and stitch them together in Clip.

Q: The generated movement does not feel like jazz dance and looks more like calisthenics. What should I do?
The characteristics of jazz dance are abrupt pauses and isolated movement in specific body parts. Write “pause on the beat” and “isolated shoulder shimmy” into the motion prompt. A more direct method is to use a standard jazz dance video as the reference for Motion Control mode.

For more features, visit NadouPro: NadouPro

Contains AI-generated information that may be incorrect. Please verify independently.