窓際お嬢様部

プロンプト

Wan 3.0 実送信プロンプト（前半・後半・福ギュンアップ） · revision 1

=== prompt_wan3_part1.txt（実送信本文） ===
Create exactly one 14-second video, segment 1 of a 41-second film.
16:9 Japanese office comedy, photoreal office and Fukuchan combined with canonical 3D characters. Exactly FOUR characters only: Sobaya, Yametaro, Fukuchan, Takosan. No Yotan or other people.
Image 1: Sobaya identity, gray skin, muscular body, white mask with red marks, fitted white T-shirt, black trousers. Image 2: Yametaro identity, compact toy proportions and purple patterned shirt. Image 3: Fukuchan exact face, black blazer, graphic white shirt, SPONSOR lanyard. Image 4: Takosan exact toy-diorama identity, black hooded robe, white face and black round eyes; two humanoid arms with rounded white fingerless mitts, six thin radial tentacles below the body. Never turn arms into tentacles.
Image 5: shared EMPTY physical set reference for BOTH segments. Preserve the exact ornate gold table and four chairs, white patterned linen, three-tier pastry stand, flower vase, teapot, immediately adjacent window mullions, city skyline, daylight and camera side. Seat Yametaro left foreground, Sobaya rear left at the beer mug, Fukuchan rear right, Takosan right foreground by the laptop. This is environment/prop reference, not an empty-scene instruction: all four canonical characters must be seated. Preserve these placements between shots and segments. Use only THREE drinking tea cups, one per non-Sobaya character; omit any extra small cup-like vessel behind the cake stand. Laptop screen must face Takosan, not the camera; maintain its position beside his place setting. Character sheets determine all faces and bodies. No sheet labels, white studio background, panels, headless bodies or duplicates.
Image 6: the previously generated same-scene ensemble, only for consistent seating, relative character sizes, table and daylight. Canonical Images 1–4 remain the identity authority. Match this layout, especially Sobaya rear-left, Fukuchan rear-right, Yametaro left foreground and Takosan right foreground.
Continuity: table directly beside and in front of large office windows, window glass and city skyline fill the background immediately behind chairs, no deep office desks behind them. Absurdly luxurious gilded afternoon tea table with white linen, silver three-tier cake stand and pastries. Normal everyday clothes, no dresses, wigs or aristocratic costumes. Left Yametaro, center-left Sobaya, center-right Fukuchan, right Takosan, all seated and never changing places. Exactly three porcelain tea cups and saucers, one each for Yametaro, Fukuchan, Takosan; exactly one large glass beer mug with golden beer and foam for Sobaya; no tea cup for Sobaya. Teapot and cake stand in center. A small open laptop near Takosan from the beginning, screen facing him with no readable UI. No actual flames or smoke: 'production is burning' means a software incident.
Sobaya keeps the opaque white mask correctly worn ALL THE TIME, including drinking and speech; never remove, lift, shift, morph or make it transparent, never expose a human face. His voice speaks through the mask; mask itself stays rigid, do not create moving human lips on it. Neck ears arms hands and all exposed skin neutral gray. White shirt preserves muscle volume and natural cloth folds, without nipple protrusions, outlines or localized paired dots/shadows.
Yametaro identity lock — Image 2 is the exact design authority, not loose inspiration. Preserve its head width-to-height ratio, hair silhouette and hairline, cheek and ear placement, short neck and compact body proportions in every frame. Keep the round black spectacle frames and the sheet's simple white lens/eye areas; do not invent visible pupils, irises, realistic eyeballs, eye sockets, eyelids or eyebrows. Keep the tiny black forehead mark separate from the low, subtle skin-colored nose; no black nose dot, outlined nostrils or projecting nose. Preserve the round cheeks and flat pink blush marks, smooth skin-colored chin, ears and nape, with no black contour bands or facial creases. Do not lengthen the forehead/head, inflate the hair, enlarge ears or add realistic lips. Animate the small inset mouth using the sheet's rest and A/I/U/E/O views as articulation guidance, with visible mouth interior when speaking; do not freeze the smile, protrude the lips or stretch the mouth into a new facial design. Emotion and cinematic lighting must not redesign the face. Wardrobe follows the scene instructions independently of identity. A previous generated take is not an identity reference.
Fukuchan's skin quality and apparent age must match the close-up portraits in his canonical character sheet. Keep his well-cared-for, clean, even complexion and natural fine skin texture. Preserve the sheet's subtle forehead lines, eye-area wrinkles, smile lines and nasolabial folds, including their natural changes with expression. Do not add or exaggerate dark spots, freckles, blemishes, rough pores or mottled skin beyond the sheet. Do not erase all wrinkles, de-age him, reshape his face, or give him plastic, waxy or porcelain skin. Preserve the scene's lighting and natural facial shading without turning shadows into blemishes.
Speech: Japanese men using elegant mock-ojousama phrasing in their canonical male voices, NOT female voices or a new narrator. One speaker at a time, no invented dialogue, no laughter track. Generate dialogue and visible mouth/jaw/cheek movement together, keep non-speakers' mouths closed, do not hide speaking mouths with cups or food. Sobaya's rigid mask is the deliberate exception to visible lip movement. Use chest-up or medium close-ups for speaking, preserve the final phoneme before cutting. Keep faces, mouths and important hand gestures above the bottom 15 percent. No on-screen text. Quiet office air and soft porcelain sounds, no background music (consistent across both segments).  All reference audio is voice-timbre only, never copy words, timing or phrasing from it.

Audio 1: sobaya, voice-timbre reference only.
Audio 2: fukuchan, voice-timbre reference only.
Audio 3: yametaro, voice-timbre reference only.
Natural speech takes priority over approximate shot timestamps. Shot ranges are editorial guides, not mandatory speech start/end times. Use natural conversational pacing with refined diction, no stretched vowels/endings or rushed delivery to hit timestamps. Finish each utterance naturally, then close the mouth. Adjust internal cuts while preserving order and total duration. Speak only the exact written dialogue. Total duration 14 seconds.
Shot 1 [0-7 s] — Begin immediately on Sobaya and his raised beer mug at the extravagantly laid table, with the adjacent window unmistakable and other companions establishing the scene. No fade or title card. Sobaya relishes slacking off, mannered and relaxed, and says exactly: 「やはり、労働をサボっていただくおビールは格別ですわね」. Mask remains fully worn and rigid. His voice is Audio 1.
Shot 2 [7-10 s] — Cut to Fukuchan, who has just taken a small bite, finishes chewing before speaking, and says brightly and casually: 「こちらのお菓子もギュンですわ」. His voice is Audio 2. Yametaro listens.
Shot 3 [10-14 s] — Cut to Yametaro holding his tea cup below his face. He casually informs Fukuchan, with matter-of-fact refined diction: 「来客用ですわ」. His voice is Audio 3. No pride, anger or special emphasis.  Then a short calm pause with a soft saucer sound. Takosan has no dialogue. End naturally at the same table, not a finale; cut to Sobaya to start the next segment. Do not start the code conversation in this segment.
Pronunciation: 労働=ろうどう; おビール=おびーる; 格別=かくべつ; ギュン=ぎゅん (one gyun, not gyungyun).
Pronunciation for Yametaro: 来客用=らいきゃくよう.


=== prompt_wan3_part2.txt（実送信本文） ===
Create exactly one 27-second video, segment 2 of a 41-second film.
16:9 Japanese office comedy, photoreal office and Fukuchan combined with canonical 3D characters. Exactly FOUR characters only: Sobaya, Yametaro, Fukuchan, Takosan. No Yotan or other people.
Image 1: Sobaya identity, gray skin, muscular body, white mask with red marks, fitted white T-shirt, black trousers. Image 2: Yametaro identity, compact toy proportions and purple patterned shirt. Image 3: Fukuchan exact face, black blazer, graphic white shirt, SPONSOR lanyard. Image 4: Takosan exact toy-diorama identity, black hooded robe, white face and black round eyes; two humanoid arms with rounded white fingerless mitts, six thin radial tentacles below the body. Never turn arms into tentacles.
Image 5: shared EMPTY physical set reference for BOTH segments. Preserve the exact ornate gold table and four chairs, white patterned linen, three-tier pastry stand, flower vase, teapot, immediately adjacent window mullions, city skyline, daylight and camera side. Seat Yametaro left foreground, Sobaya rear left at the beer mug, Fukuchan rear right, Takosan right foreground by the laptop. This is environment/prop reference, not an empty-scene instruction: all four canonical characters must be seated. Preserve these placements between shots and segments. Use only THREE drinking tea cups, one per non-Sobaya character; omit any extra small cup-like vessel behind the cake stand. Laptop screen must face Takosan, not the camera; maintain its position beside his place setting. Character sheets determine all faces and bodies. No sheet labels, white studio background, panels, headless bodies or duplicates.
Continuity: table directly beside and in front of large office windows, window glass and city skyline fill the background immediately behind chairs, no deep office desks behind them. Absurdly luxurious gilded afternoon tea table with white linen, silver three-tier cake stand and pastries. Normal everyday clothes, no dresses, wigs or aristocratic costumes. Left Yametaro, center-left Sobaya, center-right Fukuchan, right Takosan, all seated and never changing places. Exactly three porcelain tea cups and saucers, one each for Yametaro, Fukuchan, Takosan; exactly one large glass beer mug with golden beer and foam for Sobaya; no tea cup for Sobaya. Teapot and cake stand in center. A small open laptop near Takosan from the beginning, screen facing him with no readable UI. No actual flames or smoke: 'production is burning' means a software incident.
Sobaya keeps the opaque white mask correctly worn ALL THE TIME, including drinking and speech; never remove, lift, shift, morph or make it transparent, never expose a human face. His voice speaks through the mask; mask itself stays rigid, do not create moving human lips on it. Neck ears arms hands and all exposed skin neutral gray. White shirt preserves muscle volume and natural cloth folds, without nipple protrusions, outlines or localized paired dots/shadows.
Yametaro identity lock — Image 2 is the exact design authority, not loose inspiration. Preserve its head width-to-height ratio, hair silhouette and hairline, cheek and ear placement, short neck and compact body proportions in every frame. Keep the round black spectacle frames and the sheet's simple white lens/eye areas; do not invent visible pupils, irises, realistic eyeballs, eye sockets, eyelids or eyebrows. Keep the tiny black forehead mark separate from the low, subtle skin-colored nose; no black nose dot, outlined nostrils or projecting nose. Preserve the round cheeks and flat pink blush marks, smooth skin-colored chin, ears and nape, with no black contour bands or facial creases. Do not lengthen the forehead/head, inflate the hair, enlarge ears or add realistic lips. Animate the small inset mouth using the sheet's rest and A/I/U/E/O views as articulation guidance, with visible mouth interior when speaking; do not freeze the smile, protrude the lips or stretch the mouth into a new facial design. Emotion and cinematic lighting must not redesign the face. Wardrobe follows the scene instructions independently of identity. A previous generated take is not an identity reference.
Fukuchan's skin quality and apparent age must match the close-up portraits in his canonical character sheet. Keep his well-cared-for, clean, even complexion and natural fine skin texture. Preserve the sheet's subtle forehead lines, eye-area wrinkles, smile lines and nasolabial folds, including their natural changes with expression. Do not add or exaggerate dark spots, freckles, blemishes, rough pores or mottled skin beyond the sheet. Do not erase all wrinkles, de-age him, reshape his face, or give him plastic, waxy or porcelain skin. Preserve the scene's lighting and natural facial shading without turning shadows into blemishes.
Speech: Japanese men using elegant mock-ojousama phrasing in their canonical male voices, NOT female voices or a new narrator. One speaker at a time, no invented dialogue, no laughter track. Generate dialogue and visible mouth/jaw/cheek movement together, keep non-speakers' mouths closed, do not hide speaking mouths with cups or food. Sobaya's rigid mask is the deliberate exception to visible lip movement. Use chest-up or medium close-ups for speaking, preserve the final phoneme before cutting. Faces, mouths, drinks and hands above the bottom 22 percent, leaving a clean lower area for later subtitles. No generated subtitles, title, speaker labels or readable computer text. Quiet office air and soft porcelain sounds, no background music (consistent across both segments). No stingers or sounds for later subtitle joke. All reference audio is voice-timbre only, never copy words, timing or phrasing from it.

Audio 1: sobaya, voice-timbre reference only.
Audio 2: yametaro, voice-timbre reference only.
Audio 3: takosan, voice-timbre reference only.
Natural speech takes priority over approximate shot timestamps. Shot ranges are editorial guides, not mandatory speech start/end times. Use natural conversational pacing with refined diction, no stretched vowels/endings or rushed delivery to hit timestamps. Finish each utterance naturally, then close the mouth. Adjust internal cuts while preserving order and total duration. Speak only the exact written dialogue. Total duration 27 seconds.
Shot 1 [0-10 s] — Immediately begin a medium close-up of Sobaya at the same window-side tea table, mug held below mask. He casually delivers a vicious insult in refined diction: 「今度のプロジェクト、おコードがクソでしたわ。ペットのお猿にでも書かせたのかしら？」. Audio 1. Mask remains correctly worn and rigid. He does not know Yametaro wrote it.
Shot 2 [10-17 s] — Cut to Yametaro holding his tea cup below his face. He slowly turns toward Sobaya, retains a small smile but applies quiet pressure with a flat, pointed delivery, slightly stressing 全部ワイが: 「あのコードなら、全部ワイが書いてましてよ」. Audio 2. NOT proud, boasting or cheerfully volunteering; no anger shout, no altered eyes/pupils/eyelids. Body/head/voice convey 'what did you just say about me?' The subtitle joke will be added later: DO NOT turn Yametaro into a monkey or show any monkey.
Shot 3 [17-23 s] — Cut back to Sobaya, frozen for one brief beat with mug in hand, then desperately polite face-saving reversal: 「……AIには出せない、人の温もりを感じましたわ」. Audio 1. Slight hesitation then smooth praise, mask always worn.
Shot 4 [23-27 s] — One short laptop notification from Takosan's laptop, then cut to Takosan glancing at it and saying calmly, almost indifferently: 「本番もお燃えになってますわ」. Audio 3. Small mouth articulates; white mitt hands remain humanoid, not tentacles. END IMMEDIATELY when the last phoneme finishes. No closing laptop, no reaction, no group toast, no held ending, no fade or end card. If natural speech finishes before the duration limit, final edit will trim immediately after this final phoneme; never elongate the line to fill time.
Pronunciation: おコード=おこーど (source code, not cords); クソ=くそ; お猿=おさる; ワイ=わい; AI=えーあい; 温もり=ぬくもり; 本番=ほんばん (production software); お燃え=おもえ. No actual fire.


=== prompt_wan3_fukuchan_closeup.txt（実送信本文） ===
Create one continuous 4-second photoreal close-up of Fukuchan at an office afternoon tea. Image 1 is the exact canonical face, hair, skin and wardrobe authority. Image 2 supplies only the office window daylight and environment; do not show the entire table or other people.
Frame Fukuchan ALREADY close from upper chest to just above his hair, face in crisp focus and filling much of the image, centered slightly right, eyes near upper third. Fixed camera, no zoom out, no cuts. Black blazer, white graphic shirt, black lanyard as in Image 1. He sits in a gilded chair immediately beside office windows; softly defocused windows and city behind him. Ornate chair edges may appear. Tableware, pastries, laptop and other characters stay below/outside the frame. Preserve a natural photographic face, not a toy or illustration.
Fukuchan's skin quality and apparent age must match the close-up portraits in his canonical character sheet. Keep his well-cared-for, clean, even complexion and natural fine skin texture. Preserve the sheet's subtle forehead lines, eye-area wrinkles, smile lines and nasolabial folds, including their natural changes with expression. Do not add or exaggerate dark spots, freckles, blemishes, rough pores or mottled skin beyond the sheet. Do not erase all wrinkles, de-age him, reshape his face, or give him plastic, waxy or porcelain skin. Preserve the scene's lighting and natural facial shading without turning shadows into blemishes.

He looks toward a companion just left of camera, relaxed, pleased by the snack, naturally says exactly 「こちらのお菓子もギュンですわ」 once in his ordinary male voice with lightly elegant joking phrasing. ギュン is one gyun, not gyungyun. Generate natural synchronized mouth, jaw and cheek articulation. Brief calm lead-in, then a short normal conversational utterance, then close his mouth and rest with a faint smile. Do not chew while speaking. Do not elongate syllables. Audio 1 is canonical male voice timbre reference only. No music, no additional speech. No on-screen text. Keep face and mouth above bottom 15 percent. Maintain the close-up throughout.