Bottle Service
Find an accomplice
SEEDANCE 2.5 · UNIVERSAL PET SWAP PROMPT
FORMAT: 9:16 vertical, 27 seconds, one continuous shot, no cuts, 1080p, completely silent video with no audio track. STYLE: Ultra realistic live-action footage, real world physics, natural indoor daylight exactly as in @Image 2, wide-angle lens at tabletop height, shallow depth of field, high dynamic range, subtle handheld camera shake, no smooth gimbal, fine film grain, believable realism, minimal CGI appearance. REFERENCES: @Video 1 = MOTION AND CAMERA SOURCE ONLY. Copy its full-body choreography, every front-limb and paw action, head movement, timing, rhythm, framing and camera movement beat for beat. Ignore its cat's identity, fur, eyes and bandana, its living room, its table, its bottle label and its audio. @Image 1 = CHARACTER IDENTITY. This exact pet, one-to-one: same species and breed, same body shape and proportions, same fur color, length, pattern and markings, same ears, same face, same nose and whiskers, same eye color and every eye detail on the same side as in the photo. Any collar, bandana or accessory visible in the photo stays around the neck exactly as shown. Ignore the photo's floor and background. @Image 2 = ENVIRONMENT AND TABLE. This exact room and this exact table: same tabletop material, shape, edges and surface details, same furniture, walls, floor and light around it. Keep the table and room exactly as shown. VIDEO PROMPT: The pet from @Image 1 replaces the cat from @Video 1 and performs its exact movements, standing upright on its hind legs on the floor behind the table from @Image 2, using its real front paws like hands. On the tabletop: two small clear shot glasses in the middle and one larger clear tumbler close to the lens in the foreground. The bottle is a small plain white plastic bottle with a green cap and no label, filled with a white milky drink, scaled to the pet's paws. The camera sits low at tabletop level, wide lens, glasses large in the foreground, and moves exactly like the camera in @Video 1. 0.0 to 1.5s: The pet stands upright behind the table, holding the bottle out to the side in one front paw, then lifts it up beside its head. 1.5 to 3.3s: It raises the bottle high, tips its head back, brings the bottle down to chest level and gestures with its free paw. 3.3 to 6.0s: It flicks the bottle up out of the top of the frame, front paw stretched high, waiting with a playful rhythm, swaying its upper body, then catches the bottle overhead. 6.0 to 7.0s: It swings the bottle forward in a big arc and lunges toward the camera, both front paws landing on the tabletop, face now close and large in frame, every facial marking from @Image 1 clearly visible. 7.0 to 19.0s: Leaning on the table close to the lens, the pet tilts the bottle horizontally across the frame and pours a thin steady white stream into the two small shot glasses, one after the other, the glasses filling up with real liquid physics. Its eyes and head follow the pour. 19.0 to 21.0s: It moves the bottle to the large foreground tumbler and fills it. 21.0 to 22.3s: It flips the bottle up into the air above its head and rises back onto its hind legs. 22.3 to 23.0s: It catches the bottle and holds it raised in one paw. 23.0 to 25.5s: Keeping the bottle up, it leans forward and puts its other front paw down on the tabletop, slides toward the filled glasses and grips one. 25.5 to 27.0s: It pushes the full glass straight toward the lens, its paw reaching into the camera until the paw and glass fill the frame. End on this frame. AUDIO: None. Silent video. No music, no sound effects, no voice. CONSISTENCY: The pet is exactly the pet in @Image 1 in every frame: same fur color and pattern, same markings, same ears, same face, same eyes with every detail on the same side, never mirrored, never a generic look-alike. Any collar, bandana or accessory from @Image 1 stays on the whole time. Front paws are real animal paws with natural fur and soft pads. The room and the table stay exactly as in @Image 2. The bottle never leaves the pet's paws except during the two tosses. Motion, timing and camera follow @Video 1 exactly. NEGATIVE PROMPT: identity of the cat from @Video 1, bandana or accessory not in @Image 1, olive wood trunk table, glass tabletop with miniature stadium, room from @Video 1, bottle label, logo, text, wrong fur color, wrong markings, mirrored markings, eye color change, eye details on the wrong side, missing accessory, accessory color change, human hands, fingers, human body, cartoon, 3D animation look, CGI look, plastic fur, extra limbs, extra paws, deformed paws, morphing, flickering, identity drift, table change, room change, spilled liquid floating, cuts, scene change, music, sound, voice, subtitles, watermark.
Tutorial
DENIZ AKKABAK · PET SWAP TUTORIAL · @DENIZ.AKKABAK
MADE IN INVIDEO · invideo.io/i/DENIZ-ig

Bottle Service
How to hand Lil Rıfkı’s bartender routine to your own pet — one motion clip, one pet photo, one table photo, and a single Seedance 2.5 prompt that swaps everything except the moves.
A TUTORIAL BY DENIZ AKKABAK · @DENIZ.AKKABAK
PIPELINE SEEDANCE 2.5 · REFERENCE-TO-VIDEO FORMAT 9:16 · 27S · 1080P · SILENT GENRE PET SWAP / MOTION TRANSFER
The choreography is already done. Lil Rıfkı — a one-eyed black cat in a red paisley bandana — stands behind an olive-wood table, juggles a milk bottle, lunges at the lens and pours three perfect shots, all in one 27-second take. That finished take is now the reference. Seedance 2.5 can read a video as pure motion — every paw action, every head tilt, every wobble of the camera — and then recast the performer and rebuild the set around it. You do not need to reinvent a single beat: hand it the clip, a photo of your pet and a photo of your table, and the universal prompt in this guide pours the same routine with your animal behind your bar.
THE PIPELINE AT A GLANCE
- 01
The three references
@Video 1 is Rıfkı’s finished clip and carries the moves and the camera, @Image 1 carries your pet, @Image 2 carries your room and table. Each file does exactly one job.
- 02
Seedance 2.5 inside invideo
Reference-to-video, 9:16, 1080p, 27 seconds, audio off. One prompt, one continuous shot, nothing to edit afterwards.
STEP 01 — REFERENCES
Three Inputs, Three Jobs
Everything this video needs fits in three files, and the trick is that none of them overlap. The motion clip is never asked for its looks, the pet photo is never asked for its background, and the room photo is never asked for a performer. Seedance gets three clean briefs instead of one tangled one — and that separation is what lets Rıfkı’s routine survive the jump into any other animal’s body.
REFERENCE INPUTS — ATTACH IN THIS ORDER



Upload order matters. The clip goes into the reference-video slot, then the pet photo first and the room photo second, so that @Image 1 and @Image 2 in the prompt point at the right files. Choose a pet photo where the face is square to the lens — the model copies what it can read, and a three-quarter angle hides half the markings it is supposed to keep.
STEP 02 — SEEDANCE 2.5
One Prompt, One Take
Inside invideo, open Seedance 2.5 in reference-to-video mode and attach the three files. The settings mirror the source clip: 9:16 because the reference is vertical, 27 seconds because the reference runs 27 seconds — match the length and the choreography lands beat for beat instead of being squeezed or stretched. Audio generation stays off; the prompt asks for a silent take, so the soundtrack is yours to lay in the edit.
SETTINGS
| MODELSeedance 2.5 · R2V | ASPECT9:16 | RESOLUTION1080p | DURATION27 s | AUDIOOff |
SEEDANCE 2.5 · UNIVERSAL PET SWAP PROMPT
9:16 vertical, 27 seconds, one continuous shot, no cuts, 1080p, completely silent video with no audio track.
Ultra realistic live-action footage, real world physics, natural indoor daylight exactly as in @Image 2, wide-angle lens at tabletop height, shallow depth of field, high dynamic range, subtle handheld camera shake, no smooth gimbal, fine film grain, believable realism, minimal CGI appearance.
@Video 1 = MOTION AND CAMERA SOURCE ONLY. Copy its full-body choreography, every front-limb and paw action, head movement, timing, rhythm, framing and camera movement beat for beat. Ignore its cat's identity, fur, eyes and bandana, its living room, its table, its bottle label and its audio. @Image 1 = CHARACTER IDENTITY. This exact pet, one-to-one: same species and breed, same body shape and proportions, same fur color, length, pattern and markings, same ears, same face, same nose and whiskers, same eye color and every eye detail on the same side as in the photo. Any collar, bandana or accessory visible in the photo stays around the neck exactly as shown. Ignore the photo's floor and background. @Image 2 = ENVIRONMENT AND TABLE. This exact room and this exact table: same tabletop material, shape, edges and surface details, same furniture, walls, floor and light around it. Keep the table and room exactly as shown.
The pet from @Image 1 replaces the cat from @Video 1 and performs its exact movements, standing upright on its hind legs on the floor behind the table from @Image 2, using its real front paws like hands. On the tabletop: two small clear shot glasses in the middle and one larger clear tumbler close to the lens in the foreground. The bottle is a small plain white plastic bottle with a green cap and no label, filled with a white milky drink, scaled to the pet's paws. The camera sits low at tabletop level, wide lens, glasses large in the foreground, and moves exactly like the camera in @Video 1. 0.0 to 1.5s: The pet stands upright behind the table, holding the bottle out to the side in one front paw, then lifts it up beside its head. 1.5 to 3.3s: It raises the bottle high, tips its head back, brings the bottle down to chest level and gestures with its free paw. 3.3 to 6.0s: It flicks the bottle up out of the top of the frame, front paw stretched high, waiting with a playful rhythm, swaying its upper body, then catches the bottle overhead. 6.0 to 7.0s: It swings the bottle forward in a big arc and lunges toward the camera, both front paws landing on the tabletop, face now close and large in frame, every facial marking from @Image 1 clearly visible. 7.0 to 19.0s: Leaning on the table close to the lens, the pet tilts the bottle horizontally across the frame and pours a thin steady white stream into the two small shot glasses, one after the other, the glasses filling up with real liquid physics. Its eyes and head follow the pour. 19.0 to 21.0s: It moves the bottle to the large foreground tumbler and fills it. 21.0 to 22.3s: It flips the bottle up into the air above its head and rises back onto its hind legs. 22.3 to 23.0s: It catches the bottle and holds it raised in one paw. 23.0 to 25.5s: Keeping the bottle up, it leans forward and puts its other front paw down on the tabletop, slides toward the filled glasses and grips one. 25.5 to 27.0s: It pushes the full glass straight toward the lens, its paw reaching into the camera until the paw and glass fill the frame. End on this frame.
None. Silent video. No music, no sound effects, no voice.
The pet is exactly the pet in @Image 1 in every frame: same fur color and pattern, same markings, same ears, same face, same eyes with every detail on the same side, never mirrored, never a generic look-alike. Any collar, bandana or accessory from @Image 1 stays on the whole time. Front paws are real animal paws with natural fur and soft pads. The room and the table stay exactly as in @Image 2. The bottle never leaves the pet's paws except during the two tosses. Motion, timing and camera follow @Video 1 exactly.
identity of the cat from @Video 1, bandana or accessory not in @Image 1, olive wood trunk table, glass tabletop with miniature stadium, room from @Video 1, bottle label, logo, text, wrong fur color, wrong markings, mirrored markings, eye color change, eye details on the wrong side, missing accessory, accessory color change, human hands, fingers, human body, cartoon, 3D animation look, CGI look, plastic fur, extra limbs, extra paws, deformed paws, morphing, flickering, identity drift, table change, room change, spilled liquid floating, cuts, scene change, music, sound, voice, subtitles, watermark.
OUTPUT

STEP 03
Anatomy — Why the Swap Holds
- 01
The contract
FORMAT repeats what the settings already say: 9:16, 27 seconds, one continuous shot, completely silent. Redundant on purpose — when the prompt and the controls agree, the model has nothing left to negotiate, and 27 seconds matches @Video 1 to the frame.
- 02
One reference, one job
Each REFERENCES line opens with a job title in capitals — MOTION AND CAMERA SOURCE ONLY, CHARACTER IDENTITY, ENVIRONMENT AND TABLE — and closes with an ignore list. “Ignore its cat’s identity, fur, eyes and bandana, its living room, its table” is what stops Rıfkı’s face and Rıfkı’s olive-wood bar from bleeding into your shot.
- 03
Universal identity
@Image 1 never names a species or a colour; it lists what to copy — breed, proportions, fur pattern, markings and “every eye detail on the same side as in the photo.” That clause is why the same prompt works for a tabby, a husky or a one-eyed cat, and “never mirrored” in CONSISTENCY stops the model flipping a face to fit the source pose — or borrowing Rıfkı’s eyes along with his moves.
- 04
Personalise the one detail
If your pet has a signature feature, give it one sentence inside the @Image 1 line. Rıfkı’s own run spelled out his blind eye — cloudy, milky, on the cat’s left — and listed “blind eye on the wrong side” in the negatives. One precise sentence beats ten adjectives.
- 05
Rescale the props
A cat-sized bottle in a big dog’s paws reads like a pill bottle. “Scaled to the pet’s paws” resizes the prop to the new performer, and the glasses are placed by name — two shots in the middle, one tumbler near the lens — so the pour has targets that read at tabletop height.
- 06
Ten beats on the clock
The model is told to copy @Video 1, but the timecodes are the checkpoints it can be held to: the 3.3s toss, the 6.0s lunge, the twelve-second pour from 7.0 to 19.0s, the second toss at 21.0s — and “End on this frame,” so the paw in the lens is the final image, not a drift.
- 07
The insurance
NEGATIVE PROMPT bans what the reference keeps trying to smuggle in: the identity of the cat from @Video 1, any bandana your pet does not wear, the olive-wood table, the glass top with its stadium, the room. “Bottle label” is banned too — but the bottle in @Video 1 carries one, and labels love to survive. That is the honest limit of motion transfer: the more a prop moves in @Video 1, the harder it clings to its original look.
THE RESULT
Output — Lil Rıfkı Pours Three

Check it against the prompt: the bottle leaves the top of the frame and comes back, both paws hit the tabletop at the lunge, the two shots fill one after the other before the tumbler, the bandana never comes off, the cloudy eye stays on the cat’s left all the way through — and the last frame is nothing but black paw and full glass. This is the clip your run inherits, so judge your own output by the same list. If it drifts, the fix is almost always in @Image 1: a sharper, more frontal photo of your pet.
ONE MOTION CLIP · ANY PET · ANY TABLE
The moves are borrowed. The bartender is yours.
Swap the black cat for a golden retriever, the olive-wood table for a kitchen counter, the milk for whatever your pet would pour — the skeleton carries any pet into any room. @Video 1 keeps the rhythm, @Image 1 brings the face, @Image 2 builds the bar, and the timecodes make sure the last glass lands in the lens exactly where Rıfkı put his.

Seedance 2.5 lives inside invideo — reference-to-video with up to thirty image, ten video and ten audio references in a single generation, 4 to 30 seconds, up to 1080p. Attach the clip, the pet and the room, paste the prompt, and pour.
WRITTEN, DIRECTED & PROMPTED BY DENIZ AKKABAK · @DENIZ.AKKABAK