Halloween Transformation Clip — Open Skill
Halloween Transformation Clip
Q: Ask. What does this do? A: Get. I made this to turn an adult photo into a short Halloween clip. Give me a scene idea, and the costume, setting, and atmosphere change around you in one shot. Q: Ask. What do you need from me? A: Get. One clear photo with the face visible works best. Add a description with the Halloween look, location, and mood you want. Both are required. Q: Ask. Which format should I expect? A: Get. It comes out portrait, 9:16 by default. I like a crisp photo and a plain description. After-hours details usually make the scene easier to read.
Roleplay
0 0 1
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/550336/2026-10-18/sn947dvw.gif
# Halloween Transformation Clip
Pick any Halloween idea. The person starts as themselves on a moonlit street; a wipe of bats, mist or pumpkin sparks crosses them, and they emerge fully transformed in a poster-grade reveal.
## When to use
- "Turn me into a vampire / werewolf / witch / …" video with a photo
- Halloween costume reveal or glow-up reel
- Another take on a previous result with a different idea, motif or length
Not for: demon or hellfire reveals (a dedicated demon skill exists), still images only, or group scenes.
## Inputs and gates
| Input | Required | Default | Gate |
|---|---|---|---|
| Photo | Yes | — | One adult, face clearly visible. Otherwise ask for a clearer photo |
| Idea | No | `vampire noble` | Any costume or creature. If it is vague ("something scary"), choose from the idea table and say which one you picked |
| Reveal motif | No | Auto from the idea | `bat swarm`, `rolling mist`, `pumpkin sparks`, `ghost wisps`, `full-moon flash` |
| Duration | No | `6` s | 6 or 10 |
| Orientation | No | `9:16` | `16:9` on request |
- If the person may be under 18, decline.
- Keep costumes fully clothed and non-sexual. Video does not allow NSFW content.
- If the idea names a trademarked character, translate it into an unbranded look ("caped vigilante in dark armor") and say so briefly.
## Idea table (tuned defaults)
| Idea | Costume and features | Set | Motif |
|---|---|---|---|
| vampire noble | high-collar black and crimson velvet coat, pale porcelain skin, subtle fangs, red irises | candlelit castle balcony, full moon | bat swarm |
| werewolf | wolf ears, fur along the jaw and neck, amber eyes, torn flannel | misty pine forest, full moon | full-moon flash |
| witch | wide-brim hat, layered black lace robes, glowing rune pendant | foggy lantern-lit village square | pumpkin sparks |
| zombie | cracked grey skin, cloudy eyes, tattered dress or suit (stylized, no gore) | abandoned carnival at night | rolling mist |
| skeleton | Day-of-the-Dead sugar-skull face paint, black suit or gown with marigolds | candle-lined graveyard | ghost wisps |
| pumpkin king | ornate orange and black regal coat, vine crown, jack-o'-lantern glow on the face | pumpkin patch with floating lanterns | pumpkin sparks |
| ghost bride | translucent white veil and gown, pale skin, glowing eyes | moonlit chapel ruins | ghost wisps |
## Steps
Read `references/halloween-reveal.md` before you write the prompts.
1. **Identity read**: `recognize_image(image_url=<photo>, prompt='Describe this adult for identity preservation: face shape, eyes, brows, nose, lips, skin tone, apparent age range, hair, distinctive marks, glasses. One compact line.')`.
2. **Before frame**: `generate_image(prompt=<BEFORE>, input_urls=[photo], model='i2i.hd', orientation=<o>, aspect_ratio=<ar>, name='Halloween before', entity_type='work_in_progress')`.
3. **After frame**: `generate_image(prompt=<AFTER>, input_urls=[before frame, photo], model='i2i.hd', same size, name='Halloween after', entity_type='work_in_progress')`.
4. **Self-check, at most one retry per frame**: if identity drifts, put the identity clause first and calm the face effects. If the result is a timid costume, push the silhouette, materials and set.
5. **Reveal video**: `generate_video(prompt=<MOTION>, input_urls=[before frame, after frame], model='fl2v', orientation=<o>, aspect_ratio=<ar>, duration=<d>, name='Halloween Transformation', entity_type='final_videos')`. If that fails, use `model='i2v'` with `input_urls=[before frame]` and `<MOTION_I2V>`.
6. Reply in one or two sentences, naming the idea and suggesting one other idea from the table.
## Prompt templates
**BEFORE**
```text
The same adult from the reference: <identity line>. Keep their real face, hair and everyday outfit. They stand on <set> at night in the same spot the reveal will happen, soft moonlight rim, a few hints of <motif> in the background. Medium-full shot, eye-level, curious expression toward camera. Photoreal cinematic still.