Back to video
videoseedance-sceneseedance-2.0
EP.2 貞子
ホラー映画的な貞子 (cat-edition)。井戸からテレビへ。
Prompt
使用プロンプト
# 🎬 Seedance 2.0 (Runway) — 動画生成 使い方ガイド ## TOOL **Runway** (https://app.runwayml.com/) ## モデル / モード - モデル: `Seedance 2.0` - 入力: text-to-video または image-to-video ## Runway 画面の入力箇所 → 何を入れるか | Runway 画面 | このページのどこを使うか | |---|---| | **Reference Image (@IMG_1)** | 上記「Reference Image」の画像をドラッグ&ドロップ | | **Prompt** | 下記プロンプト本文をコピー | | **Duration** | 5s / 10s / 15s から選択 | | **Aspect Ratio** | 16:9 横長 (本編) / 9:16 縦長 (SNS Reels) | | **Audio** | ON 必須 (OFF だとナレーション・効果音が入らない) | | **Generate** | 60〜180秒で完成、Download MP4 | ## 注意 - 顔が映る image-to-video は Seedance のフィルタによりブロックされる場合あり → text-to-video に切り替えるか、Kling v3 を使用 ## 実投入するプロンプト本文 ```text The same Japanese man from the reference image, early 30s, keeping his exact face bones, skin tone, and facial features identical to the reference — transformed into a Japanese J-horror apparition in the style of Ringu / Ju-On. Long wet jet-black straight hair draped heavily forward, partially covering one side of his face, strands dripping with water. Pale ashen grey-tinged skin with subtle dark circles around the eyes. Wearing a soaked white Japanese burial kimono (kyokatabira) with wet folds clinging to his shoulders. Hollow haunted staring eyes visible through the parted hair, dead expression, mouth slightly open in a silent breath. He is climbing out of a dark stone well in a misty nighttime Japanese countryside. His pale wet hand grips the moss-covered stone rim in the foreground. Thick cold mist rolls across the scene, faint moonlight casts a cold bluish-white rim light on his wet hair and shoulders. Dark bamboo grove silhouettes in the distance, a single crow perched on a bare branch. Atmospheric haze, shallow depth of field. Vertical 9:16 portrait framing. Medium close-up showing head, shoulders, and hand on well rim. Photorealistic ARRI Alexa cinematic look, subsurface skin scattering, visible skin pores under the pallor makeup, 35mm film grain, desaturated cold color palette, Ringu / Ju-On aesthetic, no over-grading. Preserve the reference person's identity completely — this is clearly the same face, just transformed with wet hair and pale makeup. No English text, no Japanese text overlays, no logos, no watermarks, no cartoon, no CGI look, no exaggerated gore, no real blood. ```