My Pet Again · مهارة Open Skill عامة
My Pet Again
# My Pet Again ### First-Person POV · Single Continuous Shot · Ref2V Native > Bring back the feeling of touching your pet again. **My Pet Again** is an AI Agent Skill that transforms one animal photo into an emotional first-person interaction video. Using a reference-locked video pipeline, it creates a realistic moment where **the camera becomes your eyes**, your hand reaches into the frame, and the animal naturally responds. **One photo → Identity lock → First-person interaction → One continuous petting moment** No storyboards. No third-person shots. No artificial animal showcase. Just one intimate five-second memory-like experience. --- # Core Features | Feature | Description | |---|---| | **First-Person Pet Interaction** | Creates a realistic POV experience where you reach out and touch the animal | | **Identity Preservation** | Maintains breed, fur patterns, eyes, body shape, and unique characteristics from the uploaded photo | | **Single Continuous Shot** | One uninterrupted handheld-style clip with no cuts or camera switching | | **Emotion-First Experience** | Designed around the feeling of reconnecting with an animal, not simply showing movement | | **Universal Animal Support** | Works with pets and wildlife including cats, dogs, rabbits, tigers, lions, and more | | **Natural Animal Response** | Generates reactions based on animal type and personality | --- # Production Pipeline | Stage | Output | |---|---| | **1. Input Collection** | Upload one animal photo + optional personality description | | **2. Identity Lock** | Analyzes breed, fur, markings, eyes, body size, and visual characteristics | | **3. Interaction Design** | Creates natural petting behavior based on animal type and temperament | | **4. POV Video Generation** | Builds first-person hand interaction using ref2v technology | | **5. Final Render** | Outputs one continuous portrait interaction video | --- # Interaction System | Element | Description | |---|---| | Camera Perspective | First-person POV — camera represents the owner's eyes | | Hand Interaction | Human hand naturally enters from the bottom of the frame | | Animal Behavior | Responds with realistic movements and emotional reactions | | Shot Style | Handheld, intimate, real-life recording feeling | | Video Structure | Single continuous shot without cuts | --- # Vocalization Sync System ### Every animal sound is connected to visible behavior. | Rule | Description | |---|---| | Mouth Movement Sync | Vocalization must match visible mouth opening | | Natural Timing | Sound begins when the mouth opens and ends when it closes | | Real Behavior | Each vocalization is treated as an actual animal action | | Quiet Moments | Animals can remain silent with natural breathing, purring, or resting behavior | Example: ``` Sound begins ↓ Mouth opens naturally ↓ Jaw movement matches vocalization ↓ Sound ends ↓ Mouth closes ``` --- # Production Technology | Parameter | Specification | |---|---| | Generation Engine | Ref2V / Reference-to-Video | | Video Style | First-person POV interaction | | Duration | 5 seconds (recommended) | | Supported Duration | 5–8 seconds | | Orientation | Portrait 9:16 | | Resolution | 720P | | Camera Style | Single continuous handheld shot | | Identity Method | Upload image locked before generation | | Audio System | Visual vocalization synchronization | --- # Creation Modes | Mode | User Input | AI Output | |---|---|---| | **Photo → Video** | One animal photo | Identity lock → POV petting video | | **Personality Mode** | Photo + personality description | Creates reactions matching temperament | | **Memorial Mode** | "I want to touch my pet again" | Generates emotional reunion experience | | **Any Animal Mode** | Any animal image | Creates a first-person interaction scene | --- # Example Commands | User Request | AI Action | |---|---| | "Make a POV video where I pet my cat." | Creates a first-person cat interaction | | "I want to touch my dog again." | Creates a memorial-style reunion scene | | "Make this tiger feel like I'm gently touching it." | Generates a wildlife POV interaction | | "Make my rabbit react naturally when I pet it." | Creates personality-based response | --- # Perfect For | Use Case | Description | |---|---| | **Pet Memories** | Recreate the feeling of touching a beloved animal | | **Emotional Content** | Create heartfelt short videos for social platforms | | **Pet Owners** | Transform static photos into living moments | | **Memorial Experiences** | Create a meaningful visual memory | | **AI Creative Projects** | Generate immersive first-person animal interactions | --- # Best Results Require | Requirement | Reason | |---|---| | Clear animal photo | Improves identity preservation | | Single animal image | Keeps interaction focused | | Visible face and body | Helps maintain accurate appearance | Not recommended for: - Multi-animal scenes - Third-person animal action videos - Multi-shot cinematic storytelling - Complex scripted scenes --- # Invoke ``` /my_pet_again ``` or simply: > "Turn this photo into a first-person video of me petting it." ---
Demo
1 1 6
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/3656250/2026-08-27/9t2wrnbu.gif
# My Pet Again
## When to use
- The user uploads a photo of an animal (cat / dog / rabbit / tiger / lion, etc., **including any wild animal**) and wants to generate a realistic **first-person POV petting / interacting** video.
- The user wants to simulate the POV experience of "sitting in front of the animal, reaching out to pet it, and the animal responding naturally."
- The user mentions memorial-style animal interaction videos such as "the moment I get to touch my departed pet again" or "the first time I held it."
- **Generate directly for any animal — do not ask the user or refuse just because the animal is a wild animal / not a pet.** Whatever animal the user uploads, proceed directly with that animal.
## Core selling point
It is not about making the animal move on its own, but about **first-person intimate interaction**: the camera = the owner's eyes, a real hand reaches out from the bottom of the frame to pet the animal, and the animal responds naturally (squinting, rubbing against the hand, wagging its tail, leaning toward the camera). The emotional value is far higher than ordinary animal animation.
## Highest-priority rule: vocalization = open mouth (mandatory binding)
**Any active vocalization (especially a cat's "meow") must be bound to mouth movement as one inseparable action.** This rule takes priority over all other descriptions.
**"Closed-mouth vocalization" is absolutely forbidden**, for example: the cat's mouth is closed but a clear meow is audible, there is only an audio vocalization with no visible mouth-opening action, the mouth only moves once but the vocalization continues, or the mouth is already closed but the sound is still going.
### Mandatory synchronized event (Meow Event)
```
Meow begins → the cat's mouth starts to open
Meow continues → the cat's mouth stays open with natural jaw/mouth movement
Meow ends → the cat's mouth naturally closes
```
**The meow and the mouth opening must happen at the same time, one-to-one, fully synchronized in time.**
### Meow frequency control
Do not make the cat meow frequently just to add sound. Arrange **only one clear, brief, visible meow** in the entire video. One truly synchronized meow is far better than several consecutive unsynchronized vocalizations.
### Behavioral reason for the meow
The meow must be placed at the moment the **animal responds after being touched by the hand**, for example:
```
hand pets cat → cat looks toward camera → cat opens mouth and meows → cat closes mouth → cat leans into hand
```
## Steps
### 1. Collect input
- Animal photo (required, 1)
- Animal type (**any animal**: cat / dog / rabbit / tiger / lion, etc., can be judged from the photo)
- Animal personality (optional): gentle / playful / clingy / aloof / lively
### 2. Identify the animal's identity (lock consistency)
Call `recognize_image(image_url=<animal photo>, prompt="Describe this animal's appearance in concise bullet points: species, breed, fur color, markings/patterns, eye color, ear shape, face shape, body size/proportions, age feel.")`.
Record: breed, fur color, markings, eye color, face shape, body size, age feel. These will be written into the video prompt to lock the animal's identity.
### 3. Assemble and generate the POV interaction video
Call `generate_video`:
- `model='ref2v'`, `gen_mode='reference_to_video'` (both must be passed together to keep the animal's identity consistent)
- `input_urls=[animal photo]`
- `orientation='portrait'` (9:16)
- `resolution='720P'`
- `duration=5` (5–8 seconds)
- `prompt`: assemble per the "Video Prompt Template" below, filling in the animal type, animal identity, personality-based reaction, and the `{vocalization_sync}` snippet generated in step 3.5
- `name='Animal POV Interaction Video'`, `entity_type='final_videos'`
### 3.5 Vocalization sync system (mandatory binding)
Before generating the final prompt, first judge whether an **active vocalization event** exists per `references/vocalization-sync.md`, and generate the `{vo