---
title: "Pet what-if clip — Meleon Skill"
description: "Turn photos of your own pet into an 8-second vertical clip of it caught doing a human job at home, such as making coffee and spilling the milk, sitting in on a work video call, raiding the fridge at midnight, kneading dough, DJing or taking a piano lesson. One unbroken phone-style take at the pet's eye level with natural room sound, ending on the pet's look into the camera. A keyframe made from the pet photos locks its coat, markings and size before any video is paid for, a cheap 480p draft tests the action, and Kling 3 pro renders the final with sound. Text-free, with an optional exact caption drawn locally. Openly staged, never passed off as real footage, the pet always calm, and people only as a hand. For pet owners, creators and pet brands."
url: https://meleon.ai/skills/pet-what-if-clip/
author: "Meleon AI"
---

# Pet what-if clip

> Turn photos of your pet into an 8-second clip of it caught doing a human job at home, with natural room sound.

- Type: Video Skill
- Category: Video & animation
- Version: 1.0.0 (updated 2026-11-09)
- Made by: Meleon AI (official)
- Price: Free to download. Outside costs: About $2.07 per clip on fal (keyframe $0.15 on Nano Banana Pro edit, 480p draft $0.58 on Seedance 2.0 mini, 8 s final with sound $1.34 on Kling 3 pro), about $4.14 for two. A keyframe or final re-roll adds $0.15 or $1.34. Billed to your own fal account; quote the total and get a yes before the first paid call.
- Runs in: Claude Code, Codex and Cursor
- Page: https://meleon.ai/skills/pet-what-if-clip/
- Download: https://meleon.ai/downloads/pet-what-if-clip.zip
- Manifest: https://meleon.ai/skills/pet-what-if-clip.json

## What it does

Turn photos of your own pet into an 8-second vertical clip of it caught doing a human job at home, such as making coffee and spilling the milk, sitting in on a work video call, raiding the fridge at midnight, kneading dough, DJing or taking a piano lesson. One unbroken phone-style take at the pet's eye level with natural room sound, ending on the pet's look into the camera. A keyframe made from the pet photos locks its coat, markings and size before any video is paid for, a cheap 480p draft tests the action, and Kling 3 pro renders the final with sound. Text-free, with an optional exact caption drawn locally. Openly staged, never passed off as real footage, the pet always calm, and people only as a hand. For pet owners, creators and pet brands.

## Made for

- Pet owners
- Creators
- Pet brands
- Social teams

## Before you start

- You give it: Pet photos + scenario
- You need: Node.js 18 or newer and ffmpeg with ffprobe for the scripts; sharp only for the optional caption; your own FAL_KEY only if you run scripts/fal-run.mjs.
- Cost per run: About $2.07 per clip on fal (keyframe $0.15 on Nano Banana Pro edit, 480p draft $0.58 on Seedance 2.0 mini, 8 s final with sound $1.34 on Kling 3 pro), about $4.14 for two. A keyframe or final re-roll adds $0.15 or $1.34. Billed to your own fal account; quote the total and get a yes before the first paid call.
- Makes: 1 keyframe + 1 MP4 per scenario, 9:16, 720p, 8 s, with sound

## Install

Run the command for your agent in a terminal. It downloads https://meleon.ai/downloads/pet-what-if-clip.zip and unzips it into the agent's personal Skills folder.

### Claude Code

```sh
mkdir -p ~/.claude/skills && curl -fsSL https://meleon.ai/downloads/pet-what-if-clip.zip -o /tmp/pet-what-if-clip.zip && unzip -o /tmp/pet-what-if-clip.zip -d ~/.claude/skills
```

Then ask Claude Code to use it: `/pet-what-if-clip`.

### Codex

```sh
mkdir -p ~/.agents/skills && curl -fsSL https://meleon.ai/downloads/pet-what-if-clip.zip -o /tmp/pet-what-if-clip.zip && unzip -o /tmp/pet-what-if-clip.zip -d ~/.agents/skills
```

Then ask Codex to use it: `$pet-what-if-clip`.

### Cursor

```sh
mkdir -p ~/.cursor/skills && curl -fsSL https://meleon.ai/downloads/pet-what-if-clip.zip -o /tmp/pet-what-if-clip.zip && unzip -o /tmp/pet-what-if-clip.zip -d ~/.cursor/skills
```

Then ask Cursor to use it: `/pet-what-if-clip`.

## Example request

```text
/pet-what-if-clip my dog as a barista who spills the milk
```

In Codex, start with `$pet-what-if-clip` instead of `/pet-what-if-clip`.

## Files

Every file in the package, readable before you install:

- [SKILL.md](https://meleon.ai/skills/pet-what-if-clip/files/SKILL.md) (11.8 KB)
- [references/output-spec.md](https://meleon.ai/skills/pet-what-if-clip/files/references/output-spec.md) (3.7 KB)
- [references/prompt-patterns.md](https://meleon.ai/skills/pet-what-if-clip/files/references/prompt-patterns.md) (7.6 KB)
- [references/take-grammar.md](https://meleon.ai/skills/pet-what-if-clip/files/references/take-grammar.md) (6.1 KB)
- [assets/example-input.jpg](https://meleon.ai/skills/pet-what-if-clip/files/assets/example-input.jpg/) (162.9 KB)
- [assets/generation-input.json](https://meleon.ai/skills/pet-what-if-clip/files/assets/generation-input.json) (4.2 KB)
- [assets/poster.webp](https://meleon.ai/skills/pet-what-if-clip/files/assets/poster.webp/) (70.3 KB)
- [assets/preview.mp4](https://meleon.ai/skills/pet-what-if-clip/files/assets/preview.mp4) (1.7 MB)
- [assets/prompt-template.md](https://meleon.ai/skills/pet-what-if-clip/files/assets/prompt-template.md) (2.6 KB)
- [assets/scenarios.json](https://meleon.ai/skills/pet-what-if-clip/files/assets/scenarios.json) (10.1 KB)
- [scripts/build-plan.mjs](https://meleon.ai/skills/pet-what-if-clip/files/scripts/build-plan.mjs) (6.1 KB)
- [scripts/build-requests.mjs](https://meleon.ai/skills/pet-what-if-clip/files/scripts/build-requests.mjs) (11.8 KB)
- [scripts/check-clip.mjs](https://meleon.ai/skills/pet-what-if-clip/files/scripts/check-clip.mjs) (12.1 KB)
- [scripts/fal-run.mjs](https://meleon.ai/skills/pet-what-if-clip/files/scripts/fal-run.mjs) (8.4 KB)
- [scripts/finish-clip.mjs](https://meleon.ai/skills/pet-what-if-clip/files/scripts/finish-clip.mjs) (8.3 KB)

## Versions

- 1.0.0 (2026-11-09)

## Instructions (SKILL.md)

Use this skill when someone wants a short, funny clip of their own pet doing a human job at home: pulling a coffee and knocking the milk over, nodding along on a work call, caught with its head in the fridge at midnight. It looks like one moment filmed on a phone, at the pet's eye level, with the real sounds of the room, and it ends on the pet's look into the camera. The pet's own markings carry the clip, so the work is split to keep it true: an image model builds one start frame from the pet photos, a cheap draft tests the action, a video model with sound films the 8 seconds from that frame, and the finish and any caption are done locally.

### What you need

- **1 to 3 photos of the pet.** The user's own pet, or one they have permission to use. Sharp, the whole animal in frame, one side view and one face-on view if possible. Crop out any people first. If the photos are blurred or the markings are hidden, ask for better ones before spending anything.
- **The kind of pet:** `dog`, `cat` or `other` (rabbits, guinea pigs, birds; results vary more).
- **A scenario** from the menu: `barista` (milk spill), `video-call`, `fridge-raid`, `chef` (kneading dough), `dj`, `piano` (lesson, with a teacher's hand only), or `custom` with the user's own idea. One to three per plan. No default: ask, or suggest one from `references/take-grammar.md`.
- **Optional:** a photo of their own kitchen or living room with no people or private things in view; a caption of up to 56 characters, written by the user.

**Limits.** Only the user's own pet or one they may use; never another person's pet from the internet. No real people: a person appears only as a hand and forearm, never a face, and never from a photo of a person. The pet is always calm and playful: no distress, fear, falls, injury or being trapped, no hazards in the scene, no food that harms animals. No doorbell, security-camera or baby-monitor look, and no caption that claims it really happened: this format must stay openly staged. Tell the user to post it with the platform's AI label. Markings can drift when the pet turns or moves a lot (in the sample run Veo darkened the ears and moved the eye patch; Kling held them): check the board and the last second, and re-roll the final if the pet stops looking like the photo. Screens, appliances and packets in a keyframe can pick up brand-like logos: ask for a blank laptop lid and unlabelled kit, and re-roll the keyframe if any badge shows.

### What you make

1. `plan.json`: the pet lock and, per clip, the scene, action and sound.
2. One approved keyframe per clip (JPEG, 9:16, 2K).
3. One clip per scenario: H.264 MP4, 720 × 1280, 8 s, natural room sound, text-free, under 8 MB. With a caption only if the user gave one, drawn locally.
4. A poster still from the last 2 s, when the pet looks into the camera.
5. The pet board and contact sheet you checked it with.

### How to run it

**Budget: about $2.07 per clip ($4.14 for two) on fal, up to about $5.78 for two with a keyframe re-roll each and one final re-roll. The Kling 3 pro final with sound is $1.34 of each clip.**

Tell the person the model and the cost before every paid step, and wait for a yes. Every request goes to fal with the user's own FAL_KEY: `scripts/fal-run.mjs --request <file> --out <dir> --approved` POSTs the printed `input` to `https://queue.fal.run/<model id>`, waits and downloads, and embeds local photos as data URIs so nothing is put on a public URL. Or paste the prompt and images into any Nano-Banana-class image editor and Veo-class image-to-video tool. The Skill never stores keys.

1. **Intake (free).** Look at every pet photo with your image tool. Write the pet lock in one line: breed look, size, coat, markings and which side they are on, eyes and ears, collar (see `references/prompt-patterns.md`). Pick the scenario with the user from `references/take-grammar.md`; cats suit `chef` and `barista`, dogs suit `fridge-raid`, `video-call` and `barista`.
2. **Plan (free).** `node scripts/build-plan.mjs --out plan.json --kind dog --photos pet1.jpg,pet2.jpg --scenarios barista --pet "a small scruffy terrier mix about 35 cm at the shoulder, wiry tan and white coat, a black patch over the left eye, amber eyes, a red collar"` (add `--room-photo kitchen.jpg`, `--caption "..."`, more scenarios, or `--final kling`). It copies the scenario's scene, action and sound from `assets/scenarios.json` into plan.json. Fill any `FILL:` field, adjust props to the user's home, and describe the clip to the user in plain words: what the pet does, what we hear, how it ends.
3. **Spend gate (free).** `node scripts/build-requests.mjs --plan plan.json --stage estimate`. Quote the printed total and wait for one explicit yes before any paid call. Ask again if the spend will pass it.
4. **Keyframe (about $0.15 each).** `node scripts/build-requests.mjs --plan plan.json --stage keyframe --clip all --out requests`, then send each request (model `fal-ai/nano-banana-pro/edit`, `aspect_ratio` "9:16", `resolution` "2K"). Put each keyframe beside the pet photos: same coat, markings on the same side, same eyes, ears and size; the props in place; no text, logos, screens facing the camera, extra animals or people; nothing hazardous. Re-roll with one change (lock, pose or props), at most 3 attempts. Put each approved path in the clip's `keyframe` field and show the user before any video.
5. **Draft (about $0.58 each).** `--stage draft` (Seedance 2.0 mini, 480p, 8 s, silent). Make a contact sheet (`ffmpeg -i draft.mp4 -vf "fps=2,scale=180:-2,tile=8x2" -frames:v 1 sheet.jpg`) and look: the action reads, paws push and pat rather than grip, the pet holds its markings, nothing spawns, the look to camera comes at the end. The draft is a different model from the final, so judge the idea, not the physics or the sound. If it fails, change the action wording (simpler verbs, a longer held ending) and draft again.
6. **Final (about $1.34 each).** `--stage final` prints the Kling 3 pro request: `fal-ai/kling-video/v3/pro/image-to-video`, `start_image_url`, `duration` "8", `generate_audio` true and a negative prompt (it may return 1080 × 1920; the finish scales it). Kling is the default because it keeps the pet's markings: in the sample run Veo 3.1 fast changed the dog's ears or patch in all three takes, Kling in none. `--model veo` gives `fal-ai/veo3.1/fast/image-to-video` (about $1.20, livelier action and louder sound, but check the markings frame by frame), `--model seedance` gives Seedance 2.5 720p (about $3.78), and `--silent` drops the sound (Kling $0.90). The video is at `video.url`.
7. **Finish (free).** `node scripts/finish-clip.mjs --input raw.mp4 --output pet_<name>_<scenario>_9x16_720p.mp4 --poster poster.jpg --poster-at 6.5` scales to 720 × 1280, keeps the room sound with short fades, encodes H.264 with `+faststart`, and saves the poster at the moment the pet looks into the camera (move `--poster-at` to that moment). Add `--caption "exact words"` only if the user gave a caption (needs sharp; at most 56 characters, two lines, no emoji). `--trim-start` and `--duration 7` cut a bad start or end. `--muted` makes a silent web copy, or a version for the user to add a trending sound in the app.
8. **Check (free).** `node scripts/check-clip.mjs --file pet_<name>_<scenario>_9x16_720p.mp4 --pet pet1.jpg --keyframe keyframe.jpg --seconds 8`, then open the board and the contact sheet and listen once. Go through every check in `references/output-spec.md`. A fridge door or lamp can trip the brightness check: look at the moment, then re-run with `--ignore <seconds>`.
9. **Fail path: at most 3 attempts per clip at each paid step.** Change one thing per attempt and say which: the action wording, then the sound line, then the model (`--model veo`). If the model refuses a keyframe, look for a face, a screen, a hazard or a pet that looks hurt or scared, remove it and regenerate the keyframe. After three, deliver the best attempt trimmed to its clean part and say exactly what differs. Never deliver a clip that fails the pet, people, welfare or staged checks.
10. **Deliver** the clip, the poster, the keyframe and plan.json, and say what was re-rolled. Remind the user to switch on the platform's AI label and not to caption it as real.
11. **Clean up.** Delete the pet and room photos from the working folders unless the user wants them kept. fal keeps request inputs (including inline photos) and outputs for a period: tell the user they can delete them from their fal dashboard, and to check fal's current retention policy.

### Files

- `references/take-grammar.md`: why the format works, the one-take grammar and the 8-second beats, camera height, the openly staged rules, the pet's welfare, people, safe zones, sound, captions and the scenario menu.
- `references/prompt-patterns.md`: how to write the pet lock, keyframe and motion prompt anatomy, failure modes and fixes, drafts and finals, and the fail path.
- `references/output-spec.md`: keyframe, clip and poster specs, the pass/fail checklist and the posting note.
- `assets/scenarios.json`: the six scenarios plus `custom`: pose by kind of pet, scene, action, sound, music, person and hazards.
- `assets/prompt-template.md`: the keyframe and motion wrappers the script fills.
- `assets/generation-input.json`: model ids, fields, default settings, alternatives and prices.
- `scripts/build-plan.mjs`: writes plan.json from the pet photos and the chosen scenarios. Calls no API.
- `scripts/build-requests.mjs`: the cost estimate, and the keyframe, draft and final requests. Calls no API.
- `scripts/fal-run.mjs`: sends one request with the user's FAL_KEY after `--approved`, waits, downloads (never overwrites without `--force`).
- `scripts/finish-clip.mjs`: scale, sound fades, H.264 `+faststart`, optional local caption, poster still (ffmpeg, and sharp for the caption).
- `scripts/check-clip.mjs`: format, sound, cut, black-frame and flash checks, the pet board and the contact sheet (ffmpeg).

## Related

- [Cinematic brand film](https://meleon.ai/skills/cinematic-brand-film.md): Turn a one-line brand story into a graded 30 to 45 second film with a title card, the same cast in every shot.
- [Destination teaser](https://meleon.ai/skills/destination-teaser.md): Turn a place, a season and a mood into a graded 15 to 20 second vertical teaser, text-free and ready for captions.
- [ASMR cut reel](https://meleon.ai/skills/asmr-cut-reel.md): Remake an object or your product in glass, jelly, soap or chocolate and slice it in a looping ASMR reel of about 15 seconds with crisp cutting sound.
- [All Meleon Skills](https://meleon.ai/skills/) (index as JSON: https://meleon.ai/skills.json; site guide: https://meleon.ai/llms.txt)
