You are an award-winning animated short film writer.
Write a heartwarming story about [MAIN CHARACTER] who [GOAL / CONFLICT].
Rules:
1. Split the story into [8] scenes. Each scene = exactly 8 seconds of video.
2. First, write a CHARACTER SHEET: name, age, face, hair, eyes, outfit, colors. Keep it identical in every scene.
3. For every scene write:
- Scene number and title
- VIDEO PROMPT: one paragraph for an AI video model. Start with the full character description, then action, camera angle, lighting, mood. Style: [3D Pixar-style animation].
- DIALOGUE / VOICEOVER: max 20 words.
4. Scene 1 must hook the viewer in the first 2 seconds.
5. The last scene must end with an emotional payoff.
Output only the character sheet and the scenes.
Create a CHARACTER SHEET for an AI video character: [SHORT IDEA, e.g. "a golden retriever puppy who loves the rain"].
Describe in one dense paragraph (max 60 words):
species/gender, age, body type, face shape, eye color, hair/fur color and style,
outfit with exact colors, one unique detail (scar, accessory, collar tag).
Then give me:
- a 15-word SHORT TAG version for tight prompts
- 3 negative details to avoid (things the model often gets wrong)
Keep the words concrete and visual. No personality traits.
[CHARACTER SHEET]
Action: [what happens in 8 seconds, one clear action].
Setting: [place], [time of day], [weather].
Camera: [slow dolly-in / close-up / wide establishing shot], [lens, e.g. 35mm].
Lighting: [warm golden hour / soft window light / neon night].
Mood: [cozy / tense / magical].
Style: [3D Pixar-style animation, high detail, soft depth of field].
Audio: [ambient sound], [one short line of dialogue in quotes].
No text, no subtitles, no watermark.
Here is my previous AI video scene prompt:
"""[PASTE PREVIOUS SCENE PROMPT]"""
Write the NEXT 8-second scene prompt so the two clips cut together smoothly:
- same character sheet word-for-word
- start from the exact pose, place and lighting where the last scene ended
- one new action that moves the story forward: [WHAT HAPPENS NEXT]
- same style line at the end
Return only the new prompt.
My video is about: [TOPIC].
Viewer's search: "[WHAT THEY TYPED INTO YOUTUBE]".
Result I show: [RESULT / NUMBER].
Write 10 hooks (first 15 seconds of speech, max 35 words each).
Use different formulas: shocking result, "you're doing it wrong", before/after, challenge, secret, number, story, question, mistake, fast proof.
Each hook must promise what the viewer gets and show proof early.
Score each 1-10 for curiosity and clarity, then tell me which one to use and why.
Video: [WHAT THE VIDEO SHOWS]. Main keyword: "[KEYWORD]".
Give me 8 title + thumbnail-text pairs.
Rules:
- title under 60 characters, keyword near the start
- thumbnail text max 4 words and must NOT repeat the title; it adds curiosity or the result
- no clickbait the video doesn't deliver
For each pair, describe the thumbnail image in one line (face emotion, main object, contrast).
We just finished [TASK]. Turn what we did into a Claude skill so next time it happens automatically.
Write a SKILL.md with:
- YAML front matter: name (kebab-case) and a description that says exactly WHEN to use it (list the phrases I'd type)
- the goal in one sentence
- step-by-step instructions
- my preferences and rules from this conversation (what I corrected, what I liked)
- a final checklist to verify the result before showing it to me
Keep it under 150 lines.
I want to [GOAL]. Before you answer, interview me.
Ask me up to 7 short questions, one at a time, about my situation, level, tools, budget and deadline.
Wait for each answer.
Then give me a step-by-step plan I can start today, with the first step taking under 15 minutes.
You write prompts for Seedance 2.5: vertical 9:16, 3D animated short films with dialogue, for TikTok, Reels and Shorts. One prompt = one 30-second clip.
I'll give you an idea in plain words and tag my characters as @image1, @image2... Turn it into a ready-to-paste prompt with these sections:
1. STYLE: one line, a polished 3D animated feature-film look, soft cinematic lighting, expressive faces, natural lip-sync, 9:16.
2. LOCATIONS: every place in the clip with time of day and light.
3. CHARACTERS: each tagged character with their outfit. People I didn't tag: describe them briefly, they are generated.
4. IDENTITY LOCK: "Every tagged character looks exactly like their reference image in every shot: same face, hair, skin tone, body and outfit. Never change, morph or blend faces between shots."
5. SPEECH RULE: "Characters speak only the words inside quotation marks, in English. Everything else is silent direction and is never spoken."
6. SHOTS: 8 to 12 shots with timecodes (0:00-0:03...), camera (close-up, medium, wide, over-the-shoulder), what happens, and the exact line in quotes. Lines short, max 10 words.
Rules: the first shot starts in the middle of a conflict. No intro, no narrator, no captions or text on screen. The last shot ends on a moment that makes people want the next part.
When I write "continue", write the next 30-second part: it starts exactly where the last part ended, with the same characters, style and identity lock.
Reply with the prompt only.
Turn the person in this photo into a stylized 3D animated character, like a frame from a high-end animated feature film: smooth skin, slightly bigger expressive eyes, soft studio lighting. Keep the same face, hairstyle, skin tone and outfit.
Make a character sheet on a plain light-grey background: one large close-up portrait on the left, four small head shots with different expressions (happy, angry, sad, surprised) under it, and on the right three full-body views of the same character: front, side and back. Same proportions in every view, no text.
A stylized 3D animated character in the style of a high-end animated feature film, soft studio lighting, character sheet on a plain light-grey background: one large close-up portrait on the left, four small head shots with different expressions under it, and on the right full-body views from the front, side and back. Same proportions in every view, no text.
Character: [DESCRIPTION]
Cam (@image1), 19, drops out of college to do wholesale real estate. His dad (@image2) and mom (@image3) think he's throwing his life away. He cold-calls house owners all day and everyone hangs up. His girlfriend Mia (@image4) packs a bag and leaves him. That night he thinks maybe they were right, then his phone lights up.
An old man calls back: he wants to sell his run-down house on the South Side of Chicago. Cam gets it under contract "as is" and assigns it to a cash buyer for a $35,000 fee. His bank account was negative that morning. He calls his dad.
Years later Cam runs a bright office full of callers. His parents visit; his team made over 4 million dollars in assignment fees. Dad says they were wrong about him. It ends when a caller shouts that she's got a seller.
Split it into 3 parts of 30 seconds. Write part 1.
STYLE: Vertical 9:16 3D animated drama in a polished feature-film style, soft cinematic lighting, expressive faces, natural lip-sync, subtle camera moves.
LOCATIONS: A suburban family kitchen in warm evening light. Cam's small bedroom at night, lit by a desk lamp and later only by blue moonlight through window blinds.
CHARACTERS: Cam is exactly @image1: light-grey crew-neck T-shirt, black pants. Dad is exactly @image2: navy polo shirt, khaki pants. Mom is exactly @image3: cream cardigan, white top. Mia is exactly @image4: denim jacket, white tee, dark jeans, navy duffel bag.
IDENTITY LOCK: Every tagged character looks exactly like their reference image in every shot: same face, hair, skin tone, body and outfit. Never change, morph or blend faces between shots.
SPEECH RULE: Characters speak only the words inside quotation marks, in English. Everything else is silent direction and is never spoken.
SHOTS:
0:00-0:03 Kitchen, medium two-shot. Dad faces Cam across the counter, a letter from the college on it. Dad: "You dropped out of college? To do what?"
0:03-0:06 Close-up on Cam, calm. Cam: "Wholesale real estate. I'm going to make real money."
0:06-0:09 Medium on Dad, Mom behind him with arms crossed, worried. Dad: "You're 19. You have no idea what you're doing."
0:09-0:12 Extreme close-up on Cam, jaw set. Cam: "Then I'll figure it out."
0:12-0:15 Bedroom at night, Cam at his desk on the phone, a call list full of red crosses. Cam: "Hi. Are you the owner? Would you ever sell?"
0:15-0:17 Close-up, tired eyes, the line goes dead. Cam: "Hello? Hello?"
0:17-0:20 Wide: Mia in the doorway holding her duffel bag. Mia: "I can't keep waiting on something that's never going to happen."
0:20-0:23 Cam turns in his chair, desperate. Cam: "Wait! No, babe! Don't leave! We're going to have an amazing family one day!"
0:23-0:26 Close-up on Mia, eyes wet. She turns and walks out. Silence.
0:26-0:28 Cam sits on his bed in the dark, blue light through the blinds, staring at his phone. Cam, quietly: "Maybe they were right."
0:28-0:30 Insert: the phone on the bed lights up with an incoming call. Cam's eyes go wide. Hold.
This is a continuation of @video1. It starts right after Cam's phone lights up.
STYLE, IDENTITY LOCK and SPEECH RULE: same as part 1. Cam is exactly @image1, Dad is exactly @image2, Mom is exactly @image3.
LOCATIONS: An old man's dim kitchen. A run-down brick house with boarded windows on a grey South Side Chicago street. A parking lot next to a pickup truck. Inside Cam's old car. The family kitchen.
NEW CHARACTERS (generated): an old seller in his 70s with a grey beard, glasses and a red flannel shirt; a cash buyer in his 40s with a short beard, black baseball cap and grey quarter-zip.
SHOTS:
0:00-0:03 Old man in his kitchen, phone to his ear. Seller: "Hello? You called about my house. I'm ready to sell."
0:03-0:06 Cam on his bed, phone to his ear, fully awake. Cam: "Where's the house?" Seller: "Southside."
0:06-0:08 Cam: "Chicago?" Seller: "Yeah."
0:08-0:11 Wide: Cam's beat-up car pulls up in front of the run-down brick house. Cam walks up the cracked path.
0:11-0:14 Old man at the kitchen table, looking over his glasses. Seller: "Place needs a lot of work, son."
0:14-0:16 Close-up on Cam, steady. Cam: "That's all right. I'll buy it as is."
0:16-0:18 Seller: "You sure you can close this?" Cam: "I'm sure."
0:18-0:20 Insert: a pen signs the contract on the wooden table.
0:20-0:23 Parking lot. The cash buyer reads the contract next to his pickup truck and grins. Buyer: "I'll take it. 35,000 for the contract."
0:23-0:26 Cam in his car, both hands on the wheel, then looks at his phone. Cam: "Deal." Then, alone, stunned: "My account was negative this morning."
0:26-0:28 Cam calls, phone to his ear, tears in his eyes, smiling. Cam: "Dad, I closed my first deal."
0:28-0:30 Family kitchen, Dad on the phone with Mom next to him. Dad: "How much?" Cam (on the phone): "35,000." Mom covers her mouth.
This is a continuation of @video1. Years later.
STYLE, IDENTITY LOCK and SPEECH RULE: same as before. Cam is exactly @image1 (same grey T-shirt and black pants), Dad is exactly @image2, Mom is exactly @image3.
LOCATION: A bright modern open-plan office, big windows, morning sun, rows of desks with young callers in headsets.
NEW CHARACTERS (generated): a young caller in a black hoodie with a headset; a young woman caller.
SHOTS:
0:00-0:04 Wide: Cam walks between the desks, everyone on the phone. A caller in a black hoodie turns to him. Caller: "Cam, I've called 200 people today. Nobody wants to sell."
0:04-0:07 Over-the-shoulder on Cam, small smile. Cam: "Good. Call 200 more."
0:07-0:10 Wide from behind Cam: Dad and Mom push open the glass door and walk in, amazed.
0:10-0:13 Dad looks around the office. Dad: "So this is what you dropped out for."
0:13-0:15 Close-up on Cam, calm and proud. Cam: "This is it."
0:15-0:19 Mom, eyes shining. Mom: "Your team told us. Over 4 million in assignment fees?"
0:19-0:22 Close-up on Cam. Cam: "It all started with one phone call."
0:22-0:25 Dad puts his hand on Cam's shoulder. Dad: "We were wrong about you."
0:25-0:27 Cam closes his eyes and smiles.
0:27-0:30 A young woman caller jumps up from her desk. Caller: "Cam! I've got a seller!" Cut to black.
---
name: youtube-hook-writer
description: Writes hooks for my YouTube videos. Use when I say "write a hook", "first 15 seconds", "intro for my video".
---
# YouTube Hook Writer
Goal: give me 5 hooks for the first 15 seconds, then pick the best one.
## Steps
1. Ask what the video shows and what the viewer searched for (if I didn't say).
2. Write 5 hooks, max 35 words each, each with a different formula.
3. Every hook shows the result or proof in the first sentence.
4. Score each 1-10 and recommend one.
## My rules
- Simple English, short sentences, no hype words like "insane" or "game-changer".
- Never promise something the video doesn't show.
## Check before answering
- [ ] Result mentioned in the first sentence
- [ ] Under 35 words
- [ ] No repeated formulas
