Grokで激しめのヒップホップダンス動画を作りました!
------------------------------
画像: NanoBananaPro
動画: Grok
------------------------------
🎥 プロンプトの意識したポイント 🎥
1. 💃 ダンスの工夫:
• 単なるヒップホップではなく「insanely powerful K-POP girl-crush」と定義。
• さらに「aggressive chest pops」「confident hair flips」など具体的な動作を指定することで、この衣装に似合う、強くてセクシーな振付にしました。
2. 🎬 カット割の工夫:
• 「Dynamic multi-angle sequence: transitioning from low-angle to high-angle, snapping to close-up」と、カメラの動きをローアングル→ハイアングル→クローズアップの順番に指示するようにしました!
フルプロンプトはこちら👇
The exact same girl from the reference image — 100% identical face, long blonde straight hair with bangs, yellow cropped buckled leather jacket, black latex crop top, black leather pants with thigh cutouts, body proportions, no changes whatsoever, clean idol skin with subtle glow.
She performs insanely powerful K-POP girl-crush hip-hop choreography: aggressive chest pops, confident hair flips, deep squats, and sharp arm isolations, exuding a deadly sexy aura like a BLACKPINK main dancer.
Dynamic multi-angle sequence: transitioning from a powerful low-angle floor shot to a sweeping high-angle drone shot, occasionally snapping to a handheld close-up for impact.
Cinematic color grading: high-contrast industrial lighting, deep cool blue background smoke contrasting with her vibrant yellow jacket, bold shadows, subtle film grain, high-end music video style.
4K, photorealistic, ultra-sharp skin and fabric details, rhythmic speed ramps synced to imaginary beat, 24fps
Kling2.6でネオン街を滑空する悪魔系女子の動画を作ってみました!
やっぱKlingさん空間認識能力高そう。
距離が離れると少し人物破綻があるけど、炎のエフェクトとかいい感じ!🔥
------------------------------
画像:Midjourney, NanoBananaPro
動画:Kling2.6 Audio
------------------------------
動画プロンプトはこちら👇
cinematic action sequence of a cybernetic winged protagonist wearing a streamlined deep-purple mechanical flight suit with devil-like wings emitting black flame.
Scene 1 (vertical dive):
The protagonist launches from a skyscraper rooftop and dives head-first straight down.
Body remains aerodynamic and aligned; wings folded for maximum descent speed.
Camera follows from above and behind as neon buildings rise past in a clean vertical perspective.
Scene 2 (transition):
Moments before reaching the ground, the character opens the wings and redirects momentum into forward glide.
Orientation stays head-first with no rotation.
Environment remains stable and undistorted.
Scene 3 (horizontal flight):
The protagonist accelerates through cyberpunk streets only a few meters above the ground.
Camera tracks close behind.
Ground shows controlled motion blur while buildings stay sharp.
Black-flame trails form behind the wings, neon reflections sliding across the flight suit.
Seedance Proがつくる表情が自然すぎる。よき。
@sousakuai を使用させていただきました。
プロンプトは以下です👇
"medium close-up video of Korean-Australian idol woman with blonde hair and soft bangs looking directly at camera with a radiant, joyful smile, wearing white puffer jacket with fur-trimmed hood, wind gently blowing her hair creating natural movement, she tilts her head slightly with playful charm, her eyes sparkling with happiness, facial features remain stable and clear (no distortion), bright sunny winter day with beautiful blue sky and soft bokeh lights in background, natural sunlight creating soft glow on face, warm and cheerful atmosphere, high quality portrait, cinematic depth of field, photorealistic"
Seedance ProでCCTV風の謎にお菓子が降ってくる映像を作りました!
画質ノイズ・白飛び・UI・微ブレをなんとか表現できたかな?!?
監視映像の“リアル”がそのまま土台になって、
浮遊演出や色の強さが逆に映える感じ😎
AIだからできる表現として「現実の中に非現実を入れる」を試してみております🫡
------------------------------
画像:Midjourney + NanoBananaPro
動画:Seedance Pro
------------------------------
・固定視点 / タイムスタンプ / ノイズ / 圧縮跡 →「現実」
・浮遊するスナック / 鮮やかなネオン衣装 / スローダンス →「非現実」
このギャップがSeedance Proで自然に成立する。
ただ、監視カメラやからもっと画像荒くすればよかったです。
この動画は @sousakuai から作りました。
動画プロンプトはこちら👇
"high angle CCTV security camera footage video, looking downward from a fixed surveillance viewpoint.
A person stands below the camera and occasionally looks up toward the lens with a playful, confident expression.
They wear a vibrant rainbow tie-dye crop top with neon color accents (pink, purple, cyan, yellow).
A timestamp overlay appears at the upper left corner reading "13:00:00 12/25/2025", with a "CAM 01" label directly beneath it.
Intense sunlight streams through the windows behind, creating heavy lens flare and sensor blooming that washes out portions of the frame.
Colorful snack packages (chips, candy wrappers) float weightlessly in zero gravity, catching and scattering the light.
The visual tone is lo-fi surveillance: subtle digital noise, film grain, slight desaturation overall, but the neon colors remain vivid like "Candy Gloss".
Authentic low-resolution compression artifacts reinforce the security-camera look.
Camera remains static with no movement while the subject performs minimal, subtle gestures or dance-like motions. "
The contrast between mundane CCTV realism and the magical floating objects produces a surreal atmosphere.
Christmas daytime setting with harsh mixed lighting from interior fixtures and strong natural sunlight.
Photorealistic surveillance footage infused with Y2K pop magic.
Hailuoで生成したこのカット、ゆったりとしたダンスというかムーブ、じっくり見ると凄さが分かる。
足を組み替える時の微妙な重心の前後ブレ、 股関節の可動域と脚の入れ替わりが自然で、 ゆったり踊る時の“間”まで成立してる。
速さじゃなく「遅さと滑らかさ」で魅せる。
少しスローなダンスはHailuo2.3 @Hailuo_AI が一番信頼できる🔥
(ちょっと趣旨とズレるけどお腹辺りもリアルw)
プロンプトはGrok用のテンプレとして作ったやつで組みました。Hailuoでも良きです😊
動画プロンプトはこちら👇
Prompt:
The exact same girl from the reference image — 100% identical face, hairstyle, outfit, body proportions, no changes whatsoever, clean idol skin with subtle glow. She performs intense and sexy chair dance choreography: sitting on the vintage chair, slowly crossing and uncrossing legs, provocative deep body rolls, running hands up thighs, chin lift with a sultry teasing expression, sweat glistening on skin. Continuous low angle shot from floor level, slow smooth dolly in towards legs and dolly out to full body, steady focus on her performance without rapid cuts. Cinematic color grading: Cold blue industrial background tones with dramatic warm window side-lighting, high contrast, raw cinematic look, subtle film grain, high-end music video style. 4K, photorealistic, ultra-sharp skin and fabric details, mixture of slow-motion emphasis on leg movements and real-time flow, "raw dance practice studio style" optional, 24fps
ポラロイドに写る"何か"。
Hailuo AI @Hailuo_AI Horror Film Contestの応募作品です!
枚数を重ねるごとに、それは確実に近づいてくる。
衝撃のラストをご覧ください。
今回は音にもこだわってElevenLabsで効果音を作ったりしました!
※ホラーが苦手な方はご注意ください
Title: "What Appears"
Every photo brings it closer.
A psychological horror short film created with Hailuo AI.
Don't miss the twist at the end. 🤫
#HailuoHorror
AI動画 | ナイアガラの滝からフォール
ナイアガラの滝FPVが何度やっても“落ちてくれない問題”をなんとか潰して、やっと落下っぽいものが作れました🌊
------------------------------
画像: Midjourney v7
動画: Kling2.5Turbo
------------------------------
今回のプロンプトの改善ポイントはこれ👇
・「滝の上にいる状態」と「下が見える状態」を同時に定義
・落下開始の秒数を固定(例:3.0秒でfreefall開始)
・“垂直に落ちる”を明示するために
vertical / straight down / horizon drops をセットで記述
・船体は落下開始と同時に90°前傾(物理イベント扱いさせる)
・水流方向を常に“下向き”に固定
・ドローン化を防ぐため rigidly mounted を必ず入れる
・最後は dense white mist に自然に消失させてセンシティブ回避
AI動画モデルは「水平移動」を優先しがちな印象なので、角度変化・落下の瞬間・重力方向・カメラ固定を同時に指定すると比較的安定しました!
フルプロンプトはこちら👇
Prompt:
Ultra-realistic FPV, 10s, 9:16.
GoPro HERO12 Black rigidly mounted to the bow of a small metal boat, ultra-wide lens, natural vibration.
0–2.5s:
The boat is ON TOP of Niagara Falls, still perfectly parallel to the upper water surface.
It moves rapidly toward the brink; turbulent water pulls it forward.
The full 52m vertical drop is clearly visible ahead.
Strong sunlight, heavy mist, rainbow possible, splashes on lens.
2.5–3.0s:
The bow reaches the exact lip.
Water angle steepens; horizon drops sharply.
Camera begins to tilt forward as the boat is pulled over the edge.
At exactly 3.0 seconds → **the boat begins real freefall down the actual 52m height of Niagara Falls.**
The boat rotates naturally so that its bow becomes **almost vertical**, roughly **90° relative to the water surface**, pointing straight downward.
3.0–10s:
Full vertical freefall beside the waterfall wall, accelerating realistically under gravity.
Velocity increases continuously during the descent.
Water streams rush straight DOWN on both sides; strong vertical motion blur.
Mist becomes thicker; droplets streak the lens harder from air pressure.
The boat stays rigidly attached to the camera—no spins, no drifting.
Final seconds:
Visibility fades gradually as the boat descends deeper into dense white water mist.
The boat safely disappears into mist and spray, avoiding any sensitive impact depiction.
No people, no underwater, no gore, no third-person, no fantasy elements.
タイトル:怒ってるけどめっちゃ可愛い笑顔 × ぷくっと怒り × ツインテール × 髪色そのまま × ぷんぷんポーズ × 汗なし
ぷんぷん怒ってるけど、実はちょっと笑顔の感じを出したくて(変態)トライしました!!
なんとかまぁ及第点ができたので共有です😆
<怒りぷくツインテール汗なしの作り方>
各ポイントとそれに対応するプロンプトを併記してます。
①「変えない部分」を厳密にロックする
キャラの破綻や別人化を防ぐため、顔・肌・服・光は絶対に固定。
prompt:
"Use the attached woman as the base image.
Do not change her facial features, bone structure, makeup style, skin tone,
lighting direction, clothing, background, or overall photorealistic beauty."
② 表情は“怒り×笑顔”を両立させる(ツンデレ制御)
怒ってるのに笑顔っぽいのを出すのに苦労しました。
それには、「怒りと笑顔の要素を分離して書く」ことで安定しました!
Prompt:
"Expression:
- create a cute “puffed-cheeks angry pout,” but with a much brighter and clearly visible cute smile
- cheeks pushed forward with natural muscle tension
- lips pressed into a pout while the corners lift noticeably for a stronger adorable smile
- eyebrows pulled inward to preserve the frustrated-cute emotion
- natural blush slightly enhanced"
③ 汗・謎ハイライトを完全ブロック
AIは怒り顔に勝手に汗を足す癖があるのか、意味わからん漫画みたいな汗をかいて横したので、ブロック!! ここは本気でブロック!!許さん
Prompt:
"- absolutely NO sweat, NO moisture, NO droplets, NO wet highlights"
④ 髪型は形だけ変えて、色はそのまま
ぶっちゃけこれは好みなのでよしなにやってください☺️
Prompt:
"Hairstyle:
- convert her existing hair into realistic twin-tails
- IMPORTANT: keep the original hair color, shade, highlight pattern, and texture EXACTLY as in the base image"
⑤ ぷんぷんポーズは“腕・拳・上半身”を個別に指定
単に“pun-pun pose”と書くだけでは破綻するので、 動作を分解して書くと成功率が上がります。
Prompt:
"Pose (pun-pun pose):
- raise both arms with small clenched fists near her head
- fists slightly tighter to emphasize the “pun-pun” gesture
- slightly tilt her upper body to match the playful angry expression"
あと今回の画像は背景が秋の通学路なので、それに関するプロンプトが入っていますが、なくてもいいです!
めちゃ長くなりましたが、これで「怒ってるけどめっちゃ可愛い笑顔 × ぷくっと怒り × ツインテール × 髪色そのまま × ぷんぷんポーズ × 汗なし」画像ができます!👍
フルプロンプトはご参考までにリプに置いてます!
←元画像
怒りぷくツインテール汗なし→
🧬FLESH OF GOD🧬
Created for @HailuoAI Horror Short Film Contest 🎬
[Story]
A surgeon's attempt to transform himself into the perfect life form backfires, creating a grotesque monster.
Retaining his human consciousness, he faces the horror of what he's become and chooses poison as his final escape.
81 seconds of psychological body horror exploring the tragic consequences of playing god with one's own flesh.
✨ Deep crimson × dark green palette
✨ Bioluminescent practical effects
✨ Organic transformation sequences
✨ Haunting atmospheric score
<Japanese is following>
🧬神の肉体 - FLESH OF GOD🧬
完璧な生命体になろうとした外科医が、意図しない醜い怪物へと変貌。
自我を保ったまま己の姿の恐ろしさに直面し 、服毒による最期を選ぶ。
神を気取った代償を描く81秒の心理的ボディホラー。
✨ 深紅×暗緑の補色パレット
✨ 生物発光の実用的エフェクト
✨ 有機的変身シーケンス
✨ 薄暗く不気味な環境音楽
#HailuoHorror #ShortFilm #AIFilm
AI動画 滑空アクションチャレンジ!
カラス人間を高層ビルの屋上から垂直に滑空して、地面スレスレで地面と水平飛行する動画をつくってみました。
垂直ダイブから水平飛行への“軌道変化”を
AIに正しく理解させるには、プロンプトの構造化が必須。
今回の工夫点はこちら👇
AIが混乱しないように、垂直と水平の動きをシーンとして明確に分けました。
①Scene1:垂直落下を明記
②Scene2:transition(接地直前の切り替え)を分離
③Scene3:地面スレスレの水平フライトを独立
Grok, Veo3.1, Hailuo2.3, Kling2.5で試したけど、今回は1番イメージが近かった @Kling_ai を採用しました!
動画のプロンプトはこちら👇
Prompt:
cinematic action sequence of a humanoid crow-like figure with metallic feathers and a muscular human torso.
Scene 1 (start):
The character dives vertically from a skyscraper rooftop straight down through the neon-lit night city.
The camera follows closely from above and behind with light handheld motion, capturing the wind blasting past and the city stretching rapidly upward.
Scene 2 (transition):
Just before reaching the ground, the character pulls forward sharply, smoothly transitioning from vertical freefall into horizontal high-speed flight.
No distortion in buildings or streets; environment remains stable and coherent.
Scene 3 (final):
The character flies horizontally just inches above the ground at extreme speed.
The camera stays close behind, matching the velocity while maintaining perfect focus.
The ground passes beneath with strong motion blur, but buildings and street details remain clean and non-warped.
Neon reflections streak across metallic feathers and musculature.
Volumetric atmosphere, cinematic blue-and-amber lighting, realistic physics, ultra-dynamic perspective.
Seedance1.0マルチカット
10秒ポン出し劇場
「ハルシネーション」
映像無編集、音と字幕のみ追加
slap_sequence:
- cut: 1
shot: Bird’s-Eye View
description: Two women stand facing each other in a circle; raised voices visible in body language.
- cut: 2
shot: High Angle Medium
description: Argument peaks; one woman's hand pulls back, fingers splayed.
- cut: 3
shot: Close-Up – Hand Swing
description: Fist opens into slap motion; tension in wrist and arm muscles.
- cut: 4
shot: Shoulder-Level Close
description: POV over slapping woman’s shoulder as hand connects with the other's cheek; cheek jolts, hair flies.