Falling Halloween Pumpkin Head · সর্বজনীন Open Skill

Falling Halloween Pumpkin Head

Falling Halloween Pumpkin Head Digen-এ একটি সর্বজনীন Open Skill। খুলে দেখুন এই ওয়ার্কফ্লো কী তৈরি করে।

@GG_Bone

Custom

3 0 0

https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4150474/2026-10-15/rx89ar93.jpg

# Falling Halloween Pumpkin Head

Turns one uploaded image into a short "falling Halloween pumpkin head" video. Fixed two-step pipeline: (1) image-to-image — restyle the uploaded image into a Halloween falling-pumpkin-head still; (2) image-to-video — animate that still. There is no structured project state; the still and the clip live on the canvas.

## When to use

- The user uploads an image and asks for a "falling Halloween pumpkin head" short video.
- The user wants only one step redone (only a new still, or only a new clip from the existing still).
- The user wants the Halloween mood, props or lighting adjusted while keeping the original subject.

## Input

- The source image arrives as an attachment; pass its `asset_id` (or its `s3://` URI / https URL) as the reference input. Never invent an image reference.
- If the user did not upload an image, ask for one before running any generation.

## Steps

### Step 1 — Generate the falling pumpkin head still (image-to-image)

Call `generate_image` with the uploaded image as reference:

```text
generate_image(
  model='i2i',
  input_urls=[<uploaded image: asset_id or s3:// uri>],
  prompt=<image prompt template below>,
  orientation='portrait',
  name='Falling Pumpkin Head',
  entity_type='storyboard_images'
)
```

Image prompt template (English; replace the bracketed parts with what is actually in the reference image):

> "Turn the subject of the reference image into a falling Halloween pumpkin head while keeping the subject recognizable — same identity, same overall silhouette, same color palette and key details. Replace its head with a carved jack-o'-lantern: deep orange pumpkin, glowing warm light through the carved eyes and the jagged grin, triangular lit eyes, green stem and a short curly vine still attached. The subject plummets downward through the frame, body and limbs trailing behind, clothes / hair / foliage whipped upward by the fall, thin streaks of motion blur trailing behind the figure. Halloween night atmosphere: deep purple night sky, an oversized pale full moon behind the subject, drifting fog, bare crooked tree branches sweeping past, floating embers and dust motes, scattered small lit pumpkins and candles far below, tattered cobwebs in the corners. Warm orange rim light from the carved pumpkin faces plus cool blue moonlight, high contrast, cinematic, slightly spooky but playful, highly detailed. Vertical 9:16 composition, subject centered, headroom above for the fall, clean silhouette, no text, no watermark."

Rules for the still:

- Keep the identity the user cares about: for a person, keep face shape / hair / clothing colors recognizable; for an object or animal, keep its shape and material.
- Keep the prompt in English, and make atmosphere, props and lighting all land on the Halloween theme — fog, moon, embers, cobwebs, carved pumpkins, candles.
- Do not describe real people by name, copyrighted characters, or NSFW content. The `.hd` variants do not support NSFW; keep the bare `i2i` channel if the subject is sensitive.

### Step 2 — Animate the still (image-to-video)

Feed the still from Step 1 into `generate_video`:

```text
generate_video(
  model='i2v',
  input_urls=[<still from step 1>],
  prompt=<video prompt template below>,
  orientation='portrait',
  name='Falling Pumpkin Head',
  entity_type='final_videos'
)
```

Video prompt template (English):

> "The subject keeps falling downward as a Halloween pumpkin head — the glowing eyes and grin of the carved jack-o'-lantern stay lit and flicker slightly, the stem and curly vine drag behind, clothes and hair stream upward, fog and embers rush past the camera. Camera slowly descends and follows the subject, subject centered; the pale full moon recedes above, small lit pumpkins and candles pass by below. Smooth consistent motion, stable warm-orange and cool-blue lighting, no extra limbs, no morphing, no text."

Rules for the clip:

- Video generation does not support NSFW / adult content — refuse such requ