お正月・初詣 · সর্বজনীন Open Skill

お正月・初詣

一年の、はじまりに。 お正月と初詣を、寒気と灯りと所作の順番で撮ります。 ・8拍構成 — 鳥居、石段、手水、鈴、参拝、おみくじ、絵馬、初日 ・所作の順番が構造です — 鳥居 → 手水 → 鈴 → 参拝 → おみくじ。順番を飛ばしません ・光 — 夜明け前の冷たい青から、低い暖かい日差しへ。この移ろいが作品の核です ・音 — 除夜の鐘の余韻、石段の足音、水音、賽銭の音、遠い人混み、風。鐘は一音で十分 ・お正月の符号 — 鳥居、しめ縄、鏡餅、門松、おせち、絵馬、御守り 画面内の文字はすべて日本語です。宗教的な主張はしません。実在の神社は使用しません。オリジナル制作のみです。

@omegabubble

Film

0 0 2

https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4267104/2026-10-18/nxy6qvkr.png

---

allowed-tools:
  - generate_image
  - generate_video
  - generate_music
  - mix_clips
  - recognize_video
  - download_video
---

# お正月・初詣 / New Year Shrine Visit

New Year in Japan is **cold air, warm light, and ritual order**: the gate, the queue, the bell, the omikuji.

## When to use

- The user asks for お正月, 初詣, 新年の挨拶, or a seasonal greeting video.
- The user wants a Japanese New Year brand or family film.
- The user wants a vertical New Year short for TikTok / Shorts / Reels.
- The user uploads a reference and wants an original New Year piece.

Do **not** use this skill for: religious claims or real named shrines — see Notes.

## Steps

### 0. Clarify — which ritual, and which hour

Call `write_todos` with the plan: clarify → (optional) reference analysis → choose the setting → storyboard → stills → shots → sound → edit.

Ask only what you still need:

1. **Which hour?** — 除夜の鐘 / 初詣の朝 / 昼の参拝
2. **Which ritual?** — 参拝 / おみくじ / 絵馬 / おせち
3. **With people?** — なし / 遠景の参拝者
4. **Frame?** — portrait 9:16 / landscape 16:9

Read the choice back in one Japanese line and get a yes.

### 1. Reference analysis (only if the user supplies a clip or image)

```
recognize_video(video_url=<url>, prompt='Analyse this as a Japanese New Year shrine video. Report: the light and hour, average shot length, how static the camera is, how much of the frame is a close-up, whether people are shown face-on, and the sound design. Do not describe the plot.')
```

For a YouTube / TikTok / Instagram link, resolve it first with `download_video(url=...)`.
Turn the answer into a short treatment note and read it back. Then build an **original** piece.

### 2. Storyboard — 8 beats, about 30 seconds

| # | Beat | Length | What happens |
|---|---|---|---|
| 1 | 鳥居 | 4s | A torii at dawn, cold blue light, breath |
| 2 | 石段 | 4s | Stone steps, lanterns still lit |
| 3 | 手水 | 4s | Water poured over hands, steam |
| 4 | 鈴 | 4s | The bell rope, a hand, the sound |
| 5 | 参拝 | 4s | Two hands together, from behind |
| 6 | おみくじ | 4s | A fortune slip tied to a branch |
| 7 | 絵馬 | 5s | Wooden plaques in the cold light |
| 8 | 初日 | 5s | The first sun over the ridge. Hold |

**Rule: the ritual order is the structure — never skip a step**

### 3. 日本語での作法 — what makes it Japanese

**All on-screen text is Japanese. No English anywhere.**

**お正月の符号（必ずどれかを入れる）**
- 鳥居、石段、手水舎、鈴、賽銭箱、拝殿
- おみくじ、絵馬、破魔矢、御守り
- しめ縄、鏡餅、門松、おせち、屠蘇
- 除夜の鐘、初日の出、寒気、息
- 着物、羽織、マフラー
- 紅色と白、金

**光**
Cold blue before dawn, then a warm low sun. The transition from blue to gold is the piece.

**音**
A temple bell decaying, footsteps on stone, water, coins, a crowd murmur, wind. One bell is enough.

**テロップ**
- One line only: `一年の、はじまりに。` — mincho, white, small

### 4. Stills

```
generate_image(model='t2i', prompt='Japanese shrine at dawn on New Year: a torii gate on stone steps, paper lanterns still lit, cold blue light with a faint warm glow behind the ridge, frost on the stone, no people, no text, cinematic, subtle grain', name='鳥居', entity_type='storyboard_images', orientation='portrait')
```

```
generate_image(model='i2i', input_urls=[<user asset>], prompt='close-up of fortune slips tied to a bare branch, cold morning light, shallow depth of field, no readable text, no people', name='おみくじ', entity_type='storyboard_images')
```

### 5. Shots

```
generate_video(model='i2v', input_urls=[<still asset>], prompt='almost static: breath vapour drifting, lantern light flickering slightly, a very slow camera push toward the torii, no people, no camera shake', name='初詣', entity_type='storyboard_videos', orientation='portrait')
```

Keep `orientation` identical across every shot; never write a physical model name — channels only.

### 6. Sound

```
generate_music(model='music', prompt='Japanese New Year ambience: a deep temple bell decaying slowly, wind, footsteps on stone, a distant crowd murmur, one koto note, no drums, 45 seconds', name='除夜', entity_type='music')
```

A single Japanese line through `tts.design` is o