← Prompt library

Cat Dance to Human — Google Flow Prompt

VideoPrompt + guide9:16 preview

Illustrative cover. Results vary by model and settings.

Read the full prompt

Create a 10-second, 9:16 vertical, photorealistic live-action recreation of @Video 1 using exactly two real people from the attached identity references. REFERENCE ASSIGNMENT In Google Flow, attach the video and both identity images as Ingredients. The names below are descriptive placeholders; use the actual attached assets or their tags. @Video 1 is the original cat hallway video. Use it for the environment, relative composition, camera behavior, dance rhythm, action order, and timing. @Image 1 exclusively defines PERSON A, the DANCER replacing the small grey-and-white dancing kitten farther down the hallway, toward the middle-right. @Image 2 exclusively defines PERSON B, the FILMER replacing the white cat holding the phone in the foreground-left. Keep these identities and roles separate for the entire clip. The casting follows the supplied photos; do not invent a different person. WHOLE-BODY HUMAN REPLACEMENT Both characters are fully human from the very first frame. Preserve each reference person's recognizable facial structure, apparent age, skin tone, hairstyle, facial hair if present, and visible body characteristics. Their faces should move naturally with expression and head rotation while retaining their identities. Replace the entire cat anatomy with plausible human anatomy: human head, neck, torso, arms, hands, legs and feet. Translate the cats' movement rhythm and intention into comfortable human dance and filming movements. Use realistic joint ranges, weight shifts and foot contact. Use the clothing visible in each identity reference and keep it consistent. If the reference does not show a complete outfit, complete it with simple coordinated casual clothing, dark trousers and plain shoes. Infer only the unseen details. Do not turn the cats' fur markings into skin patches or clothing patterns. SETTING AND COMPOSITION Recreate the warm domestic corridor: pale walls, wooden floor with warm reflections, visible doorways and the darker rooms in the uploaded reference. Adapt the staging to normal human scale while preserving the source's spatial relationships. Person A dances toward the middle-right, farther from the main camera. Person B films from the foreground-left. Keep Person A visible head-to-toe, with space above raised hands and around stepping feet. Person B adopts a comfortable standing or shallow-crouching filming posture, shown mostly from a three-quarter rear angle. Allow a small natural portion of their cheek and facial profile to be visible while they remain focused on filming. Person B watches Person A or the phone screen and makes only small framing adjustments. CAMERA One continuous external observer shot, looking down the corridor past Person B toward Person A. Use a modest low-to-waist-height viewpoint adapted for human subjects, with natural perspective and enough distance to keep the full dancer in frame. Follow the source's slight early reframing, then remain steady with only subtle natural movement. Keep the spatial layout consistent. This is the external camera's view. Maintain this viewpoint for all 10 seconds. ACTION AND TIMING Use the first 10 seconds of the attached cat reference as the movement guide, at normal speed. Person A is already dancing: translate the kitten's upright arm lifts, alternating steps, small bounces and side leans into plausible human dance. Preserve the source action order and rhythm, with human joint limits and grounded feet. Person B keeps filming from the foreground-left throughout and makes only small framing adjustments. Keep one uninterrupted external-camera take; do not turn this into a duet or invent physical contact. PHONE AND SCREEN CONTINUITY Person B holds exactly one normal-size smartphone upright in portrait orientation with a secure human grip. Its rear camera points toward Person A. Its display faces Person B and is partly visible to the external camera. Whenever readable, the display shows a small coherent live rear-camera view of the same human Person A, in the same outfit and corresponding dance movement. Keep the preview consistent with the phone's viewing angle. Use a simple recording indicator without elaborate readable text. Maintain the phone's shape, scale, grip and orientation throughout. VISUAL AND AUDIO DIRECTION Natural candid home-video realism: believable skin texture, stable facial detail, soft indoor lighting, grounded shadows, natural motion blur and realistic fabric movement. Preserve the playful mood of the reference without exaggerated cartoon expressions. Follow the source video's visual timing at normal speed. Generate only subtle indoor ambience and light footstep sounds. Add no spoken dialogue, singing, meowing or new background music; the original soundtrack can be added separately during editing. AVOID Any remaining cats, including inside the phone display; cat ears, tails, fur or paws; human faces pasted onto animal bodies; identity blending or role swaps; extra people or duplicate phones; warped hands, fused fingers, sliding feet or floating bodies; outfit changes; both people becoming dancers; invented physical contact; selfie perspective; cuts, close-up inserts, dramatic zooms or camera orbits; added captions or social-media overlays.

How to use this prompt

  1. Identify the subject, audience and output you need.
  2. Replace any [bracketed placeholders] before you paste the prompt.
  3. Run it in a compatible AI tool. Change one instruction at a time if the result misses your goal.
  4. Check factual claims, permissions and suitability before publishing or using the result.

Notes from the library

GOOGLE FLOW — CAT DANCE TO HUMAN

Google Flow kholein aur naya project banayein. Settings mein Video > Ingredients chunein. Video reference ke liye Omni model use karein, agar aapke account mein available ho. Is demo mein Omni 1.1 Flash, 9:16, 720p, 10 seconds, x1 dikhaya gaya hai. Account, region aur model ke hisaab se settings badal sakti hain.

1. Download reference video se original cat clip lein.
2. Reference video aur do clear, authorized identity photos Ingredients mein add karein.
3. Photo 1 = dancer. Photo 2 = phone se record karne wala person.
4. Copy prompt karein, Flow mein paste karein aur actual asset tags assign karein.
5. Generate karein; output ready hone par faces, hands aur phone preview check karke download karein.

The source is approximately 15 seconds, 720 x 1280, 24 fps. This Flow version asks for a 10-second adaptation, matching the duration available in the inspected editor. The first photo replaces the smaller grey-and-white dancer; the second replaces the white filming cat.

REFERENCE VIDEO
https://www.aipromptking.in/uploads/videos/cat-dance-flow-reference-20261010.mp4

If video Ingredients are unavailable, do not assume that uploading a still transfers the original dance. Select a compatible model/workflow or use a separately prepared start-frame workflow. The Seedance version is a separate prompt.

This is a prompt template. No personalized two-person output has been generated or verified. Identity and phone-screen synchronization may need iteration. Do not generate until your own reference photos are attached and the roles are correct.

Official Google Flow help:
https://support.google.com/flow/answer/16353334?hl=en

Before you use the result

Use a clear reference photo only when the chosen tool supports image uploads. Use photos you have permission to edit. Describe the subject, lighting and composition separately; check faces, hands, text and background details in the result. A prompt cannot guarantee that identity or small details will be preserved.

What if the output does not match my idea?

Keep the main task, simplify conflicting instructions, and adjust one detail at a time. Describe the specific problem rather than asking only for a better result.

Does this website generate the result?

No. This website provides prompt text and guidance. Copy the prompt into a compatible AI tool; that service has its own capabilities, pricing and privacy terms.

Results vary by model and settings. Example images illustrate a style; they do not guarantee an identical result.

Learn how to improve a prompt →