Collecting in Person · öffentlicher Open Skill
Collecting in Person
Collecting in Person ist ein öffentlicher Open Skill auf Digen. Öffne ihn, um zu sehen, was dieser Workflow erstellt.
Meme
4 0 2
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4226500/2026-10-04/z7jd4ywz.png
# Collecting in Person
Turn two user-supplied images into a ~12-second funny vertical video: someone sees a photo on their phone, walks offscreen, appears inside that photo, and chases its subject.
## When to use
- The user wants the "Collecting in Person" gag: a collector sees a destination selfie on a phone, walks offscreen, enters the photo, and chases the vacationing subject.
- The user supplies (or wants to supply) two images: a **collector reference** and a **destination selfie**.
## Required inputs
1. **Collector reference** — image of the person who looks at the phone and pursues the other character.
2. **Destination selfie** — selfie showing the person being chased and their surroundings.
The two characters may be the same person. Keep their roles, clothing, and appearances distinct using their respective references. Always use the current user's supplied images as inputs — never reuse or embed a previous user's likeness.
## Default caption (editable)
`When bro owes you money but posts himself on vacation.`
Ask the user if they want a different caption; otherwise use this default. Burn it in as a subtitle overlay on the final mix.
## Dialogue (editable)
Specify all spoken lines and burn them in as on-screen subtitles (spoken audio is not reliably generated). Default lines:
- Clip 1: "What the—?!" then "Bro, you owe me money and you're on vacation?!"
- Clip 2: "Wait—what?!" then "Get back here!"
Ask the user if they want different lines; otherwise use these defaults.
## Workflow
Before generating anything, briefly explain the proposed workflow and the unsupported parts (see Notes). Use `write_todos` to list the steps and update as you progress.
### 1. Build two consistent starting stills (`generate_image`, model=`i2i`, orientation=`portrait`)
- **Tabletop still** — `input_urls=[collector_ref, destination_selfie]`. Prompt: the collector is in an **indoor living room** (sofa, TV, lamp, coffee table) — a setting clearly distinct from the destination photo — leaning in from the side of the table, **looking directly at the phone screen** (reaction-video framing, not behind the phone); the phone screen shows the destination selfie scene; the collector's expression is offended disbelief; keep the collector's appearance from the collector reference and the destination scene from the destination selfie. `name='cip_tabletop_still'`, `entity_type='cip_stills'`.
- **Destination still** — `input_urls=[destination_selfie, collector_ref]`. Prompt: the destination selfie expanded into a full scene, preserving the subject's outfit, props, and recognizable surroundings; the collector now stands behind the subject in the background, approaching; the subject holds a selfie phone at arm's length. `name='cip_destination_still'`, `entity_type='cip_stills'`.
### 2. Clip 1 — living-room reaction + push-in (`generate_video`, model=`i2v`, orientation=`portrait`)
`input_urls=[cip_tabletop_still]`. Prompt: in the living room, the collector looks at the phone screen, expression changes to offended disbelief ("What the—?"), props the phone upright on the table, then walks out of frame to the left; the camera stays behind and smoothly pushes toward the phone screen; the screen fills the frame. `name='cip_clip1_tabletop'`, `entity_type='cip_clips'`.
### 3. Transition clip — photo comes alive (`generate_video`, model=`i2v`, orientation=`portrait`)
`input_urls=[cip_destination_still]`. Prompt: the destination scene begins as a frozen photograph and smoothly comes alive with continuous motion (no cuts): the collector standing behind the subject begins to move and approach. This bridges the phone-screen close-up into the live scene. `name='cip_transition'`, `entity_type='cip_clips'`.
### 4. Clip 2 — inside the photo + chase (`generate_video`, model=`ref2v`, orientation=`portrait`)
`input_urls=[cip_destination_still, collector_ref, destination_selfie]`. Prompt: the photograph becomes a moving scene; the collector appears behind the subject (offs