KAGURA | AI映像スタジオ運営

AI cinema & working prompts, daily (EN/JP)｜和風AIシネマを一人で制作（映画祭応募作あり）｜実際に動いたプロンプトを全文公開｜実費・失敗もそのまま｜Higgsfield/Seedance｜

[KAGURA | AI映像スタジオ運営 @studiokagurajp]

Start here / 入口まとめ

🎬 Edo-Japan cinematic AI films — 和風AIシネマを一人で制作中
📝 毎日、実際に動いたプロンプトを全文公開（EN/JP）

▼ Higgsfield【PR】紹介リンク（新規は3時間限定・最大50%オフ、条件あり）
https://t.co/0vlvyb7eCD

▼ プロンプト集・制作実費・手順（note）
https://t.co/PazRpmqZpp

[KAGURA | AI映像スタジオ運営 @studiokagurajp]

I wrote "bronze wind bell." Summer furin are clear glass — and "small" plus "close-up" kept making a temple bell fill the frame.

Take three: fist-sized glass, beam thicker than the bell, paper strip longer than the bell. I stopped banning glass and wrote one chirin line instead. Picture finally reads right; I'm checking if the ping matches.

「青銅の風鈴」と書いていました。語源は寺の風鐸の金属です。夏に軒下でチリンと鳴るのは、透明なガラスの方です。
それと「small」と「close-up」を同時に書くと、小さい物が画面いっぱいになります。v1もv2も、軒の梁より大きい鐘になっていました。

三回目で、拳くらいのガラス、梁はその何倍も太い、短冊は鈴より長い、と書き直しました。酒の縁と同じで、「禁止だけ」より「ここまで／この音」を一行足した方が効きました。絵はようやく風鈴に見えます。音は今夜、耳で確認します。

Made with @higgsfield / Seedance.

[KAGURA | AI映像スタジオ運営 @studiokagurajp]

3 prompt lines that stop AI cloth from sliding like silk.
AIの布が「絹みたいに滑る」のを止める、3行です。

I asked for one knot and cotton friction. The knot held. My hands stayed on the lacquer rim instead of resting beside the box — I kept it anyway.
結びは一発で締まったんです。木綿の摩擦も効いています。なのに最後、手が箱の横に置かれず、漆の縁を持ったまま止まりました。それでも残しました。

PROMPT (worked, as-is):
SCENE: Low wooden floor, 16:9. Locked top-down close-up on a plain indigo cotton wrapping cloth around a square black lacquer box. Adult hands, thirties, hold two opposite corners. Face out of frame. No writing. No pattern with text.

ACTION: One knot. The hands cross the two corners, pull once, and the knot tightens. The cloth draws in around the box. The hands let go and rest beside it. No second knot. No untying.

FOCUS: Tack-sharp on the knot, the cotton weave, and the fingertips. The floor is one stop soft.

PHYSICS: Cotton has thickness and friction. It creases where it is pulled and stays creased. The box does not move. It does not slide like silk. No hover.

LIGHT: Window light from the left. Matte cloth. A small specular on the lacquer edge only.

STYLE: Photoreal cinematic, period Japan, indigo cotton and black lacquer, soft grain. No anime. No text. No watermark. No logos.

AUDIO: Diegetic SFX only. No music. No score. No singing. No speech. No narration. One soft cotton pull and a short cloth rasp as the knot closes. Then quiet.

Model: Seedance 2.5 @ Higgsfield / 1080p / 10s / Sound On / no Start frame

The line that mattered → PHYSICS: Cotton has thickness and friction. It creases where it is pulled and stays creased. It does not slide like silk.

From the side it almost looks like tying a necktie on a box. Almost.

[KAGURA | AI映像スタジオ運営 @studiokagurajp]

@Xxx_note_xx 🪮

[KAGURA | AI映像スタジオ運営 @studiokagurajp]

I asked for sake rising just above the rim. It stopped a few millimeters short.

My guess: "No overflow" was the easiest line to obey. Stop pouring early, and the rule is kept.
Next take gets a height to reach, not just a line not to cross.

縁から少しだけ盛り上がるお酒を頼みました。
注ぎ方は綺麗なんです。徳利も、猪口も、行灯の光も。なのに真ん中に泡をちょっと残して、縁の数ミリ手前でぴたっと止まりました。

たぶん「溢れさせないで」の一行が、いちばん守りやすかったんだと思います。早めに注ぐのをやめれば、それで守れてしまうので。
なので次は「越えないで」じゃなくて、「ここまで上がって」と書いてみます。

Made with @higgsfield / Seedance.

[KAGURA | AI映像スタジオ運営 @studiokagurajp]

Soft things don't need more adjectives. They need limits.
やわらかい物に説明を足すと、だいたいゴムになります。

I gave the tofu three jobs: lift, sag once, return one drop. No wobbling. No collapse.
豆腐には三つだけ頼みました。少し持ち上がる。少したわむ。一滴戻る。豆腐にしては、まあまあ忙しい。

PROMPT (worked, as-is):
SCENE: Low table, 16:9. Locked close-up on a small square of silken tofu in a dark ceramic dish. Adult hands, thirties, hold one pair of wooden chopsticks. Face out of frame. No readable glaze mark.

ACTION: The chopsticks take the tofu from two sides and lift it a few centimeters. The block holds, then sags a little. A drop of whey falls back to the dish. No bite. No second lift.

FOCUS: Tack-sharp on the tofu skin, the chopstick tips, and the whey drop. Background table is soft.

PHYSICS: Tofu is soft and wet. It has weight. It does not turn to rubber. It does not explode. No hover.

LIGHT: Window light from the left, a small specular on the wet surface.

STYLE: Photoreal cinematic, period Japan, soy sheen and wood grain, soft grain. No anime. No text. No watermark. No logos.

Model: Seedance 2.5 @ Higgsfield / 1080p / 10s / Sound Off / no Start frame

For something soft, do you write shape or weight first?
やわらかい物なら、形と重さ、どちらを先に書きますか？

[KAGURA | AI映像スタジオ運営 @studiokagurajp]

I wrote this in 16:9. It came back wide.
16:9で書きました。戻ってきたのは、もっと横長でした。

At that angle the crossing is a little hard to place. I kept the take. The stone still clicks once, and it stays.
その角度だと、どの交点か少しわかりにくい。それでもこのテイクにしました。石は一度鳴って、そこに残ります。

Made with @higgsfield / Seedance.

[KAGURA | AI映像スタジオ運営 @studiokagurajp]

I didn't film this. I wrote the weight.
撮影していません。この石の重さを書きました。

The first version touched the board like it had no mass. I stopped describing the room and gave the stone one job: land on one intersection, then stay there.

最初は盤に触れているのに、重さがありませんでした。部屋の説明を減らして、「交点に一つ置いて、そのまま止まる」だけを書きました。

PROMPT (worked, as-is):
SCENE: Wooden go board, 16:9. Locked close-up on empty intersections, grid only. A slate-black convex go stone is held between adult fingers, thirties. A simple wooden bowl sits soft at the far edge. Face out of frame. No writing on the board. No inlay text. No plastic shine.

ACTION: One placement. The black stone is set on a single intersection and stays. The fingers withdraw a little and hold. No second stone. No white stone. No sweeping the board.

FOCUS: Tack-sharp on the stone edge, the wood grain of the point, and the fingertip skin. The rest of the board is one stop soft.

PHYSICS: The stone has weight. It clicks onto the wood and sits. It does not hover or slide into place. No magnet snap.

LIGHT: Window light from the left, a small specular on the polished wood only.

STYLE: Photoreal cinematic, period Japan, slate stone and worn wood, soft grain. No anime. No text. No watermark. No logos.

AUDIO: Diegetic SFX only. No music. No score. No singing. No speech. No narration. One stone click on wood. Then quiet. No second click.

Model: Seedance 2.5 @ Higgsfield / 1080p / 10s / Sound On / no Start frame
Cost & time: Not logged for this take.

Keep the verb small. The contact will do the rest.
動詞を小さくすると、接地が残ります。

If you use it, save the prompt before changing the subject.
使うなら、被写体を変える前にこのまま保存してください。

[KAGURA | AI映像スタジオ運営 @studiokagurajp]

I didn't film this. I wrote the weight.
撮影していません。この石の重さを書きました。

The first version touched the board like it had no mass. I stopped describing the room and gave the stone one job: land on one intersection, then stay there.

最初は盤に触れているのに、重さがありませんでした。部屋の説明を減らして、「交点に一つ置いて、そのまま止まる」だけを書きました。

PROMPT (worked, as-is):
SCENE: Wooden go board, 16:9. Locked close-up on empty intersections, grid only. A slate-black convex go stone is held between adult fingers, thirties. A simple wooden bowl sits soft at the far edge. Face out of frame. No writing on the board. No inlay text. No plastic shine.

ACTION: One placement. The black stone is set on a single intersection and stays. The fingers withdraw a little and hold. No second stone. No white stone. No sweeping the board.

FOCUS: Tack-sharp on the stone edge, the wood grain of the point, and the fingertip skin. The rest of the board is one stop soft.

PHYSICS: The stone has weight. It clicks onto the wood and sits. It does not hover or slide into place. No magnet snap.

LIGHT: Window light from the left, a small specular on the polished wood only.

STYLE: Photoreal cinematic, period Japan, slate stone and worn wood, soft grain. No anime. No text. No watermark. No logos.

AUDIO: Diegetic SFX only. No music. No score. No singing. No speech. No narration. One stone click on wood. Then quiet. No second click.

Model: Seedance 2.5 @ Higgsfield / 1080p / 10s / Sound On / no Start frame
Cost & time: Not logged for this take.

Keep the verb small. The contact will do the rest.
動詞を小さくすると、接地が残ります。

If you use it, save the prompt before changing the subject.
使うなら、被写体を変える前にこのまま保存してください。

[KAGURA | AI映像スタジオ運営 @studiokagurajp]

I thought plain gold would look unfinished.
So I almost added a painting. Then the screen became the painting, and the gold was just a frame. I took the picture out and left the window light on the folds.

金地だけだと、未完成に見えそうでした。
だから絵を足しかけました。すると屏風が絵の額縁になる。絵を外して、窓の光だけが折れ目に残るテイクにしました。

Made with @higgsfield / Seedance.

[KAGURA | AI映像スタジオ運営 @studiokagurajp]

I almost put a painting on this screen. The gold looked borrowed.
最初は屏風に絵を入れかけた。金だけだと物足りないと思った。

I took the painting out. Then the window light finally had somewhere to go.
でも絵を外したら、光が走る場所ができた。残したのはそのテイクです。

PROMPT (worked, as-is):

SCENE: A tatami room, 16:9. Locked three-quarter on a six-fold gold-leaf byobu standing in a zigzag. The gold is plain — no painted scene, no famous composition. An adult is not in frame. No readable writing.

ACTION: The gold starts dim. A cooler band of window daylight slides across the folds as if a shutter opened, then holds. Light faces and shadow faces of the folds stay in place. The screen does not walk. This is not a wipe.

FOCUS: Tack-sharp on the gold leaf grain and one fold edge at center. Far tatami is one stop soft.

PHYSICS: Gold reflects. It does not sparkle like glitter. The screen has weight and stays planted. No CGI shimmer loop.

LIGHT: One window from the left. Specular runs along the gold, then settles. No museum even-fill. No electric color.

STYLE: Photoreal cinematic, period Japan, gold leaf and lacquered frame, soft grain. No anime. No text. No watermark. No logos.

Model: Seedance 2.5 @ Higgsfield / 1080p / 15s / Sound Off / no Start frame
効いているのはこの1行 → “The gold is plain — no painted scene.” 絵を足す前に、光の居場所を残す。

Save this for your next shoot 🔖
次の生成前に見返せるよう保存どうぞ

[KAGURA | AI映像スタジオ運営 @studiokagurajp]

I almost put a painting on this screen. The gold looked borrowed.
最初は屏風に絵を入れかけた。金だけだと物足りないと思った。

I took the painting out. Then the window light finally had somewhere to go.
でも絵を外したら、光が走る場所ができた。残したのはそのテイクです。

PROMPT (worked, as-is):

SCENE: A tatami room, 16:9. Locked three-quarter on a six-fold gold-leaf byobu standing in a zigzag. The gold is plain — no painted scene, no famous composition. An adult is not in frame. No readable writing.

ACTION: The gold starts dim. A cooler band of window daylight slides across the folds as if a shutter opened, then holds. Light faces and shadow faces of the folds stay in place. The screen does not walk. This is not a wipe.

FOCUS: Tack-sharp on the gold leaf grain and one fold edge at center. Far tatami is one stop soft.

PHYSICS: Gold reflects. It does not sparkle like glitter. The screen has weight and stays planted. No CGI shimmer loop.

LIGHT: One window from the left. Specular runs along the gold, then settles. No museum even-fill. No electric color.

STYLE: Photoreal cinematic, period Japan, gold leaf and lacquered frame, soft grain. No anime. No text. No watermark. No logos.

Model: Seedance 2.5 @ Higgsfield / 1080p / 15s / Sound Off / no Start frame
効いているのはこの1行 → “The gold is plain — no painted scene.” 絵を足す前に、光の居場所を残す。

Save this for your next shoot 🔖
次の生成前に見返せるよう保存どうぞ

[KAGURA | AI映像スタジオ運営 @studiokagurajp]

The first take was prettier. It was also wrong.
最初のテイクの方が、たぶん綺麗でした。でも違った。

I had the camera follow the figure, and the street got smaller. I stopped it and let the person walk out of frame instead.
笠の人を追わせたら、夜道の奥行きが消えた。カメラを止めて、人物だけが画面の外へ出るようにした。

Once they leave, the lanterns and the empty street do the work.
人が抜けたあと、提灯と暗い通りの方が、人物より夜らしく見えた。

Made with @higgsfield / Seedance.

[KAGURA | AI映像スタジオ運営 @studiokagurajp]

Sept 24 is the Sora API cutoff. I already had a street I could keep.
9/24にSora APIが止まる。残せる夜道は、もう手元にあった。

I wrote a walk with weight. The brim never lifts. I kept the pass that exits.
歩きに重さを書いた。笠は上がらない。画面を抜けたテイクを残した。

PROMPT (worked, as-is):

SCENE: A narrow wooden street at night, 16:9. Locked camera, chest-height. Paper lanterns under eaves recede in depth. Wet packed earth, not asphalt. An adult, thirties, in a dark kimono and a deep kasa, starts at the right edge. Face hidden by the brim. No second person. No shrine gate. No readable shop signs.

ACTION: The figure walks left through the frame in one continuous pass and exits. The walk has weight. The kasa brim never lifts. Camera does not pan. No second figure. No look-back.

FOCUS: Tack-sharp on the walking figure's shoulders and the nearest lantern paper. Far eaves are one stop soft.

PHYSICS: The body has weight. Feet meet the wet ground. The person does not glide or hover. Cloth lags a little, then catches up.

LIGHT: Warm lanterns only. Cool darkness between them. No modern streetlight. No beauty key on the face — the face is not seen.

STYLE: Photoreal cinematic, period Japan, wet wood and oil-paper lanterns, soft grain. No anime. No text. No watermark. No logos.

AUDIO: Diegetic SFX only. No music. No score. No singing. No speech. No narration. Weighted footsteps on wet packed earth. Soft cloth. Night air between lanterns. The sound leaves with the figure.

Model: Seedance 2.5 @ Higgsfield / 1080p / 15s / Sound On / no Start frame
効いているのはこの1行 → PHYSICS の "does not glide or hover." 歩かせるなら、滑らせない。

Try it and show me — I read every reply.
試したらリプで見せてください。全部見ます

[KAGURA | AI映像スタジオ運営 @studiokagurajp]

Sept 24 is the Sora API cutoff. I already had a street I could keep.
9/24にSora APIが止まる。残せる夜道は、もう手元にあった。

I wrote a walk with weight. The brim never lifts. I kept the pass that exits.
歩きに重さを書いた。笠は上がらない。画面を抜けたテイクを残した。

PROMPT (worked, as-is):

SCENE: A narrow wooden street at night, 16:9. Locked camera, chest-height. Paper lanterns under eaves recede in depth. Wet packed earth, not asphalt. An adult, thirties, in a dark kimono and a deep kasa, starts at the right edge. Face hidden by the brim. No second person. No shrine gate. No readable shop signs.

ACTION: The figure walks left through the frame in one continuous pass and exits. The walk has weight. The kasa brim never lifts. Camera does not pan. No second figure. No look-back.

FOCUS: Tack-sharp on the walking figure's shoulders and the nearest lantern paper. Far eaves are one stop soft.

PHYSICS: The body has weight. Feet meet the wet ground. The person does not glide or hover. Cloth lags a little, then catches up.

LIGHT: Warm lanterns only. Cool darkness between them. No modern streetlight. No beauty key on the face — the face is not seen.

STYLE: Photoreal cinematic, period Japan, wet wood and oil-paper lanterns, soft grain. No anime. No text. No watermark. No logos.

AUDIO: Diegetic SFX only. No music. No score. No singing. No speech. No narration. Weighted footsteps on wet packed earth. Soft cloth. Night air between lanterns. The sound leaves with the figure.

Model: Seedance 2.5 @ Higgsfield / 1080p / 15s / Sound On / no Start frame
効いているのはこの1行 → PHYSICS の "does not glide or hover." 歩かせるなら、滑らせない。

Try it and show me — I read every reply.
試したらリプで見せてください。全部見ます

[KAGURA | AI映像スタジオ運営 @studiokagurajp]

I put Mio on YouTube. I made it in @higgsfield.
Mio を YouTube に上げた。@higgsfield で作った。

https://t.co/NYyEsNsbW5

[KAGURA | AI映像スタジオ運営 @studiokagurajp]

AI didn't ruin this shot. My first verb did.
ショットを壊したのはAIじゃなく、最初の動詞だった。

I wrote "fog." The alley looked like a machine. I changed it to rolls, and kept the take where the sheet thins.
「fog」と書いた。ドライアイスになった。rolls に替えて、厚くなって薄れるテイクを残した。

PROMPT (worked, as-is):

SCENE: Edo night alley, 16:9. Locked three-quarter on a wooden shoulder-yoke soba stall, no wheels. One paper andon glows. A boiling pot sits in the stall. An adult vendor, thirties, wears a kasa; the face stays hidden under the brim. No second face. No readable writing on the noren or lantern.

ACTION: Steam rolls off the pot and crosses the lens in one slow sheet, then thins. The vendor does not look up. No pour into a bowl. No customer stepping in.

FOCUS: Tack-sharp on the steam edge, the pot rim, and the andon paper. The alley boards are one stop soft.

PHYSICS: Steam is hot and wet. It drifts, then fades. It does not look like dry ice or a fog machine. No CGI ribbons.

LIGHT: The andon is the only warm key. Cool wet night around it. Specular on the broth only.

STYLE: Photoreal cinematic, period Japan, wood smoke and oil-paper glow, soft grain. No anime. No text. No watermark. No logos.

AUDIO: Diegetic SFX only. No music. No score. No singing. No speech. No narration. No vendor call. Soft boil and a wet steam hiss as the sheet crosses the lens. Distant night alley air.

Model: Seedance 2.5 @ Higgsfield / 1080p / 15s / Sound On / no Start frame
効いているのはこの1行 → ACTION の "rolls … one slow sheet, then thins." fog とも sit とも書かない。

Steal this. Swap the pot. Ship it.
鍋だけ替えてそのまま使ってOK

[KAGURA | AI映像スタジオ運営 @studiokagurajp]

I almost deleted the hang. The splash needed a miss first.
浮いてる前半、捨てかけた。着地の水の輪には、失敗が先に要った。

Same locked shot. Two beats. No second step.
カメラはそのまま。拍は二つ。二歩目は書かない。

@higgsfield / Seedance

[KAGURA | AI映像スタジオ運営 @studiokagurajp]

3 prompt lines that stop AI video from floating.
AI動画の「浮き」を止める3行

I wanted the foot to hang first. When it lands, the water ring is what sells the weight.
浮かせたあと、着地させた。水の輪が出たテイクを残した。

Grounding trick: one locked shot, two beats. Float → contact. No walk-cycle.

PROMPT (worked, as-is):

SCENE: Wet stone path at dusk, 16:9. Locked low close-up. An adult man's geta, thirties. Face out of frame. No second person. No readable writing.

ACTION: Two beats in one locked shot. First half: the geta hangs a finger's width above one dark wet cobble and stays there. No splash. No contact. Second half: the wooden sole meets the same cobble and holds. A thin water ring spreads from the contact. The strap tightens. No walk-cycle. No second step.

FOCUS: Tack-sharp on the sole edge, the wet grain of the wood, and the stone sheen. Background path is one stop soft.

PHYSICS: First half is the miss — the foot floats. Second half has weight. Water displaces. No ice-slide. No hover after contact.

LIGHT: Cool dusk from the right, a small specular on the wet stone only.

STYLE: Photoreal cinematic, period Japan, worn wood and wet stone, soft grain. No anime. No text. No watermark. No logos. No split-screen graphic. No captions.

AUDIO: Diegetic SFX only. No music. No score. No singing. No speech. No narration. Near-silence while the geta floats. On contact: one wet wood-on-stone plant and a small splash as the water ring spreads. No second step.

Model: Seedance 2.5 @ Higgsfield / 1080p / 30s / Sound On / no Start frame
効いているのはこの3行 → ACTION の "No walk-cycle. No second step." ／ PHYSICS の "No hover after contact." ／ AUDIO の contact splash。歩かせない。

Bookmark now, thank yourself at 2am.
深夜の生成中の自分のために保存を

[KAGURA | AI映像スタジオ運営 @studiokagurajp]

3 prompt lines that stop AI video from floating.
AI動画の「浮き」を止める3行

I wanted the foot to hang first. When it lands, the water ring is what sells the weight.
浮かせたあと、着地させた。水の輪が出たテイクを残した。

Grounding trick: one locked shot, two beats. Float → contact. No walk-cycle.

PROMPT (worked, as-is):

SCENE: Wet stone path at dusk, 16:9. Locked low close-up. An adult man's geta, thirties. Face out of frame. No second person. No readable writing.

ACTION: Two beats in one locked shot. First half: the geta hangs a finger's width above one dark wet cobble and stays there. No splash. No contact. Second half: the wooden sole meets the same cobble and holds. A thin water ring spreads from the contact. The strap tightens. No walk-cycle. No second step.

FOCUS: Tack-sharp on the sole edge, the wet grain of the wood, and the stone sheen. Background path is one stop soft.

PHYSICS: First half is the miss — the foot floats. Second half has weight. Water displaces. No ice-slide. No hover after contact.

LIGHT: Cool dusk from the right, a small specular on the wet stone only.

STYLE: Photoreal cinematic, period Japan, worn wood and wet stone, soft grain. No anime. No text. No watermark. No logos. No split-screen graphic. No captions.

AUDIO: Diegetic SFX only. No music. No score. No singing. No speech. No narration. Near-silence while the geta floats. On contact: one wet wood-on-stone plant and a small splash as the water ring spreads. No second step.

Model: Seedance 2.5 @ Higgsfield / 1080p / 30s / Sound On / no Start frame
効いているのはこの3行 → ACTION の "No walk-cycle. No second step." ／ PHYSICS の "No hover after contact." ／ AUDIO の contact splash。歩かせない。

Bookmark now, thank yourself at 2am.
深夜の生成中の自分のために保存を

[KAGURA | AI映像スタジオ運営 @studiokagurajp]

I didn't film this. I wrote it.
撮影していません。書きました

I wanted thunder. The garden went white, then rain hit the stones. I kept that take.
雷が欲しかった。庭が白くなって、石に雨が当たった。そのテイクを残した。

PROMPT (worked, as-is):

SCENE: A small stone garden at night, 16:9. Locked wide-close. Dark gravel, one pine, wet cobbles. No face. No shrine gate. No temple. No readable carving.

ACTION: The frame starts nearly black. One photographic lightning flash whites the garden for an instant, then darkness returns a stop brighter. Rain begins at once. Heavy drops hit the cobbles and bounce. No second flash. No person entering.

FOCUS: Tack-sharp on the wet cobbles as the rain hits. The pine is one stop soft.

PHYSICS: Lightning is a real sky flash, not a drawn zigzag. Rain has weight. Drops splash and run. They do not hang as fog. No CGI bolt lingering in the air.

LIGHT: The flash is the only hard key. After it, cool wet night. No beauty fill.

STYLE: Photoreal cinematic, period Japan, wet stone and pine needles, soft grain. No anime. No text. No watermark. No logos.

AUDIO: Diegetic SFX only. No music. No score. No singing. No speech. No narration. One thunder crack with the flash, then heavy rain hitting cobbles. No second thunder.

Model: Seedance 2.5 @ Higgsfield / 1080p / 15s / Sound On / no Start frame
効いているのはこの1行 → ACTION の "One photographic lightning flash whites the garden for an instant." 空にギザギザを描かない。二発目を足さない。

Steal this. Swap the garden. Ship it.
庭だけ替えてそのまま使ってOK

[KAGURA | AI映像スタジオ運営 @studiokagurajp]

90% of my prompt is camera. 10% is subject.
プロンプトの9割はカメラ。被写体は1割

Light changes. Wood and paper stay still.
動くのは光だけ。木と紙は止まっている。

PROMPT (worked, as-is):

SCENE: Interior facing a closed shoji, 16:9. Locked frame. The paper grid fills the shot. No hand. No silhouette of a person. No readable writing.

ACTION: The paper starts dim. A cooler daylight brightens across the panels as if a cloud has passed, then holds. The wood lattice does not move. This is not a door opening. Not a wipe.

FOCUS: Tack-sharp on the paper fiber and one kumiko joint at center. Edges stay in frame.

PHYSICS: Light changes. Wood and paper stay still. No warp. No camera drift.

LIGHT: Diffuse daylight through paper only. No electric color. No hard sun disk drawn on the paper.

STYLE: Photoreal cinematic, period Japan, worn wood and washi tooth, soft grain. No anime. No text. No watermark. No logos.

Model: Seedance 2.5 @ Higgsfield / 1080p / 10s / Sound Off / no Start frame
効いているのはこの1行 → PHYSICS の "Light changes. Wood and paper stay still." 戸を開けない。ワイプにしない。

[KAGURA | AI映像スタジオ運営 @studiokagurajp]

Mio.

A woman with a hundred faces.
One of them is hers.

百の顔を生きた女。
最後に残るのは、名前だけ。

#Higgsfield #HiggsfieldFilmFestival

[KAGURA | AI映像スタジオ運営 @studiokagurajp]

The strongest prompt edit is deleting a word.
最強のプロンプト改稿は「1語消す」

It does not become a fog wall.
霧の壁にしない。

PROMPT (worked, as-is):

SCENE: Tatami edge at night, 16:9. Locked close-up on a small ceramic incense burner. A single stick stands. No face. No readable writing on the ceramic.

ACTION: One thin thread of smoke rises, wavers once, and thins. The stick glows at the tip only. No second stick. No hand entering.

FOCUS: Tack-sharp on the ember tip and the first ten centimeters of smoke. Background fusuma is one stop soft.

PHYSICS: Smoke is light and continuous. It does not become a fog wall. It does not pool like liquid. No CGI ribbon.

LIGHT: The ember is the only warm point. Cool room around it. No beauty key.

STYLE: Photoreal cinematic, period Japan, ceramic glaze and ash texture, soft grain. No anime. No text. No watermark. No logos.

Model: Seedance 2.5 @ Higgsfield / 1080p / 10s / Sound Off / no Start frame
効いているのはこの1行 → PHYSICS の "It does not become a fog wall." 一筋だけ。二本目も手も入れない。