Bedtime Reels Comedy · Open Skill 공개 스킬
Bedtime Reels Comedy
Bedtime Reels Comedy은 Digen의 공개 Open Skill입니다. 열어 보면 이 워크플로가 무엇을 만드는지 알 수 있습니다.
Custom
0 1 2
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4239354/2026-10-19/wguxpk39.jpg
# Reels Bedtime Comedy ("Just Five More Minutes")
## When to use
- The user wants a photorealistic, cinematic, **silent**, single-shot vertical (9:16) comedy video of a person lying in bed at night scrolling their phone, falling asleep, then waking up in shock as morning light reveals the whole night passed.
- The user wants to redo or restyle this one-shot bedtime comedy clip (different character photo, tweaked timing, changed setting details, etc.).
## Character reference
- Use the user's uploaded personal photo as the **exact identity reference** for the main character (or the preset `character_photo` if one is bound). Preserve the person's facial features, hairstyle, skin tone, and overall appearance throughout the entire video.
- If no photo is provided, ask the user to upload one before generating.
## Steps
1. Confirm the character reference photo (user upload or preset `character_photo`). This is the only input image.
2. Call `generate_video` with:
- `model='ref2v'` (reference-to-video — preserves the character's identity from the photo)
- `input_urls=[<character photo asset_id or uri>]`
- `orientation='portrait'` (vertical 9:16 for Reels)
- `duration=8` (use up to 10 if the user wants a longer beat)
- `name='Reels bedtime comedy'`
- `entity_type='final_videos'`
- `prompt=` the full continuous-shot description below.
3. The video must be **completely silent** — do not add dialogue, voiceover, music, or sound effects. Do not call any audio tool.
4. Reply with the finished clip in natural language (don't paste raw URLs).
## Prompt template (single continuous shot)
"Photorealistic cinematic single continuous shot, vertical 9:16, about 8–10 seconds, completely silent. A tired man lies comfortably on his bed at night in a cozy dimly-lit bedroom with a comfortable bed, soft pillows, a blanket, and a smartphone. He holds the smartphone close to his face and scrolls through short videos; his eyes are heavy but he is still curious and entertained. He adjusts his pillow, gets more comfortable, and keeps scrolling as if he intends to watch for only five minutes. The camera slowly moves closer to his face; his eyelids grow heavier and his head sinks into the pillow, his thumb continues scrolling lazily for a moment, then he falls asleep with the phone still in his hand, expression peaceful and unaware of the time passing. Suddenly his eyes snap open wide as if realizing something terrible; he looks down at the phone, then quickly toward the window, his expression shifting from sleepy confusion to exaggerated shock and disbelief. The camera smoothly pulls back to reveal bright morning sunlight flooding the bedroom, now clearly lit by daylight; the man is still lying in the same position holding the smartphone, exhausted and stunned, realizing the entire night has passed. End on his shocked facial expression for a brief comedic beat. Natural human movement, expressive facial acting, realistic bedroom details, warm nighttime lighting transitioning naturally to bright morning sunlight, smooth camera movement, believable continuity, excellent comedic timing. The phone screen shows no readable text and no recognizable app interface. No cuts that change the man's position unexpectedly, no extra people, no distorted hands, no duplicate phones, no subtitles, no on-screen text."
## Notes
- This is a **single continuous take** — one `generate_video` call, no cutting or stitching. Do not split into scenes.
- The night-to-morning transition must be visually clear within the one shot.
- If the result has continuity problems (identity drift, distorted hands, duplicate phones, unexpected position changes), redo the generation with a tightened prompt rather than trying to patch it with cuts.
- Report failures honestly; never fabricate URLs.