From One Character Image to a 16-Panel Dance Sheet and a Continuous Dance Video: A Two-Step Prompt with GPT Image 2.5

X creator Kiki (@Mayz1169) published a two-step prompt technique on September 10, 2026 that turns a single uploaded image into a character-consistent dance reference sheet and a continuous dance video. Step one generates a 4x4, 16-panel pose sheet with GPT Image 2.5; step two composites it into one continuous animation with Seedance 2.5 on the Renoise platform.

Image source: @Mayz1169 on X

The author wrote, "GPT Image 2.5 works surprisingly well for dance references! Just upload one image and you can create a dance video like this," and shared the full prompts. Renoise (@renoiseai) described the toolchain as: "GPT Image 2.5 can turn any character into a dance sequence from just one image. Create 16 poses with GPT Image 2.5, then animate them with Seedance 2.5 on Renoise."

Step 1: The 4x4 Dance Pose Sheet Prompt

Feed one character image into GPT Image 2.5 as the reference and use the prompt below verbatim. The output is a single square image with 16 equally sized panels in a 4x4 grid.

Create a 4×4 dance pose sheet featuring the character in the uploaded reference image, with exactly 16 equally sized panels in one square image.

CHARACTER CONSISTENCY
Use the uploaded character as the sole reference for character identity and appearance. Preserve the character's face, hairstyle, colors, body proportions, species, outfit, accessories, and original art style across all 16 panels. Do not redesign the character or add an idol costume. If any body or costume details are not visible in the reference, infer them conservatively and keep them consistent throughout.

ADAPTIVE CHOREOGRAPHY
Design a cohesive dance sequence tailored to this specific character's personality, appearance, anatomy, clothing, and range of motion. Choose a suitable dance style automatically: lively for an energetic character, graceful for an elegant character, confident for a bold character, playful for a cute character, or restrained for a serious character. Do not force every character into cute idol poses.

Show 16 distinct key poses from the same dance routine, ordered from left to right and top to bottom. Make the sequence feel connected, with natural transitions and a clear beginning, build-up, climax, and ending.

Use the following movement structure, adapting each action to the character:
1. Opening pose that establishes the character's personality.
2. Small rhythmic step with a coordinated arm gesture.
3. Side step with one arm extended.
4. Compact expressive gesture near the face or upper body.
5. Grounded stance with arms crossing or sweeping inward.
6. Knee lift or an anatomically appropriate weight shift.
7. Open, expansive pose with arms spread.
8. Playful or expressive balance pose.
9. Turning pose showing a three-quarter rear view.
10. Forward-reaching movement with a diagonal body lean.
11. Strong upward-reaching accent.
12. Climactic jump or a grounded full-body extension if jumping is unsuitable.
13. Dynamic landing and recovery pose.
14. Smooth spin transition maintaining center of gravity.
15. Anticipation pose leading into the grand finale.
16. Confident ending pose freezing on the beat.

Two points matter most. First, the consistency block locks the face, hairstyle, colors, proportions, species, outfit, accessories, and original art style across all panels, forbids redesigns or added idol costumes, and tells the model to infer unseen details conservatively but consistently. Second, the choreography block picks a dance style to fit the character's personality, look, anatomy, clothing, and range of motion instead of forcing every character into cute idol poses. The 16 poses run left to right, top to bottom, with a clear arc of opening, build-up, climax, and ending.

Step 2: The Video Prompt That Stitches the Sheet into One Animation

Feed the sheet from step one into the video model as the choreography and appearance reference. The author's example describes a blonde high-side-ponytail idol character; swap the appearance details for your own character.

Use this sheet for character appearance and choreography. Follow poses 1→16 from left to right, top to bottom, connecting them into one continuous dance.

Create a continuous 2D anime animation of one cheerful idol dancing in a plain white space. Preserve her golden-blonde high side ponytail, curled hair tips, bright blue eyes, navy-white-and-gold star hair bow, small star hair clips, orange-and-white fitted stage vest, black necktie, gold trim and star ornaments, white pleated skirt with navy edging and orange ruffles, long waist ribbons, black fingerless gloves, and tall white boots with orange cuffs and gold stars. Match the reference's body proportions, crisp outlines, vibrant colors, and clean cel shading. Keep her face, hairstyle, costume, and accessories consistent. No microphone or handheld props.

Her performance is bright, playful, and confident, combining lively idol gestures, rhythmic steps, heart-shaped hands, light hops, and a triumphant finishing pose. Use natural smiles and occasional winks.

Use a fixed front-facing camera in one uninterrupted full-body shot. Start immediately in pose 1. Keep her entire silhouette, including raised fingers, ponytail, ribbons, and boot soles, inside the frame with comfortable white margins. A soft light-gray oval shadow grounds her feet and remains on the floor during jumps.

The grid defines choreography order only. Show one character in one continuous space. No reference sheet, split screens, borders, numbers, captions, duplicate figures, afterimages, cuts, or camera movement.

Each numbered pose occupies 0.5 seconds. Clearly articulate its silhouette, then transition through natural movement into the next. Maintain continuous anatomy and weight transfer: feet push off and land, knees absorb impact, and hips and shoulders guide turns. Her ponytail, skirt ruffles, and ribbons follow with slightly delayed motion. Avoid slideshow transitions, morphing, foot sliding, or unnatural pacing.

The shooting constraints are specific: a fixed front-facing camera, one uninterrupted full-body shot, starting immediately in pose 1, with the full silhouette — raised fingers, ponytail, ribbons, boot soles — kept inside the frame with comfortable white margins, and a soft light-gray oval shadow that stays on the floor even during jumps. The grid defines choreography order only: no reference sheet, split screens, borders, numbers, captions, duplicate figures, afterimages, cuts, or camera movement. Each pose holds 0.5 seconds, articulated clearly before transitioning naturally into the next, with continuous weight transfer — feet pushing off and landing, knees absorbing impact, hips and shoulders leading turns, and hair, skirt ruffles, and ribbons following with slightly delayed motion. Slideshow transitions, morphing, foot sliding, and unnatural pacing are explicitly excluded.

Model Compatibility Caveat

According to the author and the quoted source, the animation step uses Seedance 2.5 on the Renoise platform. Per user feedback (@toiren22), some video models such as MiniMax H3 misread the 16-panel grid as "16 people moving at once" rather than one sequential routine, so per-frame sequential interpolation fails. Use a model that can read the 16-panel grid in order and composite it into a single-character video (the Seedance 2.5 family), or add a frame-splitting preprocessing step.

Original source

• Kiki (@Mayz1169) original post: GPT Image 2.5 works surprisingly well for dance references!
• Renoise (@renoiseai) quoted post: GPT Image 2.5 can turn any character into a dance sequence