Independent AI fashion film · Self-initiated · 2026
A real garment. An imagined afternoon.
An independent, self-initiated film. The Summer House and its Takumi trousers were the inspiration; the brand did not commission or endorse it and was not involved in making it.
One More Minute began with a small human situation: a woman is ready to leave, but she isn’t quite finished with the afternoon. Set in an imagined Bengaluru apartment, the film follows the light, the movement of cloth and a brief moment of private play.
The experiment was to make an AI fashion film feel lived in. Character, garment, room, movement and sound had to support the same moment. This page shares the references, decisions and iterations behind it, including the full prompt for the final generation.
The human situation
She has dressed to meet a friend waiting below. Her friend calls from outside, and instead of leaving she lets the call become a rhythm. A foot answers it, then a few loose steps and a half-turn, and for a moment she is simply enjoying her own company. A second call catches her at it. She smiles, says “Coming” and goes.
That shift, from practical readiness to private play to an affectionate departure, is the whole film. It also gives the garment a reason to move: the trousers are seen in use, not posed.
Three lines, written for the film
“I was ready.”
“I just wasn’t finished with the afternoon.”
“Coming.”
Building the ingredients
Each ingredient was worked out on its own before anything moved. Casting and natural-skin studies settled on one face. The official product photographs were studied for construction: the broad waistband, where the pleats begin, the finished opening at the outer lower leg. A separate study developed the room and its light.
Three of these became inputs to the final generation, each with a single job. The room study did not; by then, the earlier cut carried the room.
Planning the moment
Before any motion was generated, the situation was laid out on an editable director canvas in Figma and drawn as a 12-panel storyboard, at a provisional 47 seconds. The boards were a scaffold, not a shot list. Four of the panels show what they settled.
All twelve panels
The order of work here, with references, boards and a director canvas before motion, was informed by a public production breakdown of the animated short Passport Rush and its project canvas on Higgsfield. That was an animation workflow; I adapted it for a fashion film. Its production assets belong to Higgsfield and are not shown here.
A study of a Calvin Klein fashion film shaped what this one should feel like: performance, intimate photography, rhythm and the human life around a garment. It was a study of published work; I had no part in the campaign, and its footage is not reproduced here.
“Sadie Sink Feels the Fit | Calvin Klein Fall 2026”, Calvin Klein. Directed and photographed by Aidan Zamiri.
The iteration that changed the film
The first aim was a film of 40 to 50 seconds. Motion was built in stages: recognition, cloth movement and departure, then a connected scene and a fuller assembly. The first full assembly ran about 40.6 seconds; a tighter cut of about 28.1 seconds kept the natural pause and had colour matching and Foley.
That cut held together, and it felt too restrained. Instead of patching shots, I used the finished cut as a reference for the world and its continuity, and generated a new 30-second sequence with Seedance 2.5. The prompt names what had to change: replace the long holds and sparse action “with a growing private performance and a clear return to life.”
Taken from the earlier cut
Left behind
The new sequence was generated at 720p, 16:9, and measured 30.093 seconds. It was later upscaled to 3200 × 1800.
The recipe
The final generation combined one video and three images, run through FLORA. Each reference had a job and a limit, and where they disagreed, the prompt let the product photographs override the clothing in the video.
2.5-5s — Cut to intimate eye-level close portrait. Eyes turn toward the window before her head; she recognizes her friend below. A slight private acknowledgment, not a beauty smile. Her internal voice, dry and conversational, says exactly: "I was ready." Her lips stay quiet during voiceover.
Recognition is written as a physical order, eyes before head, and the default beauty smile is ruled out by name. The line is internal, so her lips are told to stay quiet.
11-13s — Cut to a brief side detail continuing the last step: fabric brushes her calf; the existing side opening briefly reveals it; the two trouser legs remain separate. Light rakes across woven cotton and fine pleats. Cut while the cloth is still moving.
The garment gets its own shot, but the shot continues her step instead of stopping for a product close-up, and the cut comes while the cloth is still moving.
The timings are direction for the model, not a record of where each moment lands in the film. Generation is not deterministic: the same prompt and references will not make this film again.
Exactly as sent for the final generation. Timecodes and headings are highlighted; the wording is unchanged.
Use @Video1, the supplied finished film, as a reference for this woman's lived-in Bengaluru apartment, window-to-door geography, pale blue shirt styling, intimate camera distance and believable everyday presence. Extract the shaded window light on the left, cool mineral walls, warm living skin, matte stone floor, stool/book and doorway on the right. Create a NEW 30-second edited fashion commercial with the action and sound below. The source is continuity evidence, not an edit to duplicate: replace its long holds and sparse action with a growing private performance and a clear return to life. Discard its soundtrack, titles and inaccurate broad trouser pleats.
@Image1 is the CHARACTER authority: the same fictional adult woman, late twenties, long chestnut hair, blue-green eyes, light olive skin, recognizable facial proportions and ordinary skin variation. Preserve her identity through profile, movement and close views; her reference smile is not a permanent expression.@Image2 is the PRODUCT front-construction authority: The Summer House Takumi off-white woven cotton trousers, a continuous broad high waistband, many closely spaced narrow pleat origins releasing into two distinct wide legs, matte textured cloth and ankle-length hems.@Image3 is the PRODUCT side-opening authority: preserve the finished outer lower-leg opening, its upper join and hem. It opens with the stride and falls closed with gravity; it is not a floating flap. Product images override garment errors in @Video1. Do not inherit the product model, sandals or studio background. Our woman wears the pale blue cotton shirt partly tucked, Takumi trousers and bare feet throughout.
ONE MORE MINUTE. A woman ready to meet a friend below lets a familiar call become a rhythm, gets briefly carried away in her own company, then leaves. Observational, tactile, witty and physically alive. One woman, one room, one continuous afternoon. This is an edited sequence of distinct shots at natural speed. Each cut continues an action or changes our access to her thought. No dissolving between different bodies, repeated poses or automatic slow motion.
0-2.5s — START INSIDE AN ACTION. Tight waist-and-hand frame, normal perspective, camera just below chest height. She finishes one practical shirt tuck and releases her fingers; expose the actual waistband and dense fine pleats. A distant human two-note whistle comes through the open window at about 1.4s. Her hand stops because she hears it.2.5-5s — Cut to intimate eye-level close portrait. Eyes turn toward the window before her head; she recognizes her friend below. A slight private acknowledgment, not a beauty smile. Her internal voice, dry and conversational, says exactly: "I was ready." Her lips stay quiet during voiceover.5-7.5s — Cut low to both bare feet and lower trousers on stone. One soft foot contact, then the other, playfully answers the call's rhythm. Weight transfers through planted feet; the cotton follows a fraction later. These two contacts seed the music.7.5-11s — Cut to stable diagonal full-body view, window left and doorway right. The small rhythm becomes two loose lateral steps and a settled recovery. A shoulder joins in almost despite herself. She stays clear of the stool. Allow the viewer to see her whole silhouette and real balance.11-13s — Cut to a brief side detail continuing the last step: fabric brushes her calf; the existing side opening briefly reveals it; the two trouser legs remain separate. Light rakes across woven cotton and fine pleats. Cut while the cloth is still moving.13-16s — Medium-wide, camera follows gently sideways as she makes one easy half-turn, plants her feet and settles. Hair and trousers lag behind her torso and fall back. A small unguarded breath of laughter. Relaxed arms, no staged dance flourish.16-20s — Close portrait after movement, softly displaced hair and a little breath left in her body. A brief satisfied pause now has a cause. Internal voice says exactly: "I just wasn't finished with the afternoon." Keep this lightly amused, warm and clear, without a sales-presenter cadence. Her lips remain silent.20-24s — The friend repeats the two-note whistle from outside. Let the music drop back. Medium-wide: eyes register the interruption, she is caught enjoying herself, smiles once, turns toward the established doorway and takes the first departing step. No new pose or reset.24-27.5s — Three-quarter view along her path to the door. Two or three purposeful steps carry her out, trousers moving with her. Near 25s she says aloud, gently amused, "Coming." Natural lip sync only for this word. She clears the doorway before the ending.27.5-30s — Hold the now-empty room briefly, light and air still present. Footsteps recede beyond the door. On the quiet wall, a restrained two-line typographic finish reads exactly "THE SUMMER HOUSE" and, smaller, "TAKUMI". Stable clean letters; no other text. End on settled room tone.
PHOTOGRAPHY: intimate fashion film with physical camera positions: moderate-wide perspective for whole-body movement, normal/short-telephoto proximity for face and cloth. Keep the action axis and room geography consistent. Shaded late-afternoon window light shapes the near cheek; quieter room-bounce shadows keep dimension. Neutral off-white cotton, pale blue shirt, warm skin against subdued green-grey surroundings. Soft highlight rolloff, restrained contrast, subtle optical softness while eyes and weave remain legible. Skin has regional sheen, gentle under-eye folds and irregular fine hairs; no smoothing, pore-noise overlay or cosmetic gloss. Preserve cloth weight and contact with the body. No floating feet, duplicated limbs, altered waist construction or wardrobe change.
ORIGINAL NATIVE AUDIO: generate the entire soundtrack with the film. Quiet leafy residential Bengaluru air, very distant everyday traffic, two clearly human whistles outside, close soft cotton friction, bare-foot contacts on stone and one small exhaled laugh. An original sparse brushed-percussion and warm bass groove grows from her foot rhythm in the middle, then recedes on the second call. No sung lyrics. Voiceover is one adult woman's intimate natural English voice; audible, unhurried and lightly self-aware. Only the exact two VO sentences and the one spoken word above; no extra narration or subtitles. Keep voices clearly above music, footsteps proportionate, ambience continuous across cuts. Preserve the energy change from practical readiness to private play to affectionate departure.
What I learned
Each input in the final generation had a single responsibility and a stated limit, and the prompt says what not to take from each: the earlier cut’s garment errors, the product model, her sandals, the studio background. When references disagree, the prompt says which one wins.
The earlier cut kept the woman, the room and the garment consistent from shot to shot, and still felt restrained. Continuity is the floor, not the film.
A held moment reads as satisfaction rather than waiting when the action before it gives it a reason. The prompt puts it plainly: “A brief satisfied pause now has a cause.”
The prompt treats skin as one problem: shaded window light on the near cheek, regional sheen, gentle under-eye folds and fine hairs, no smoothing, and expressions that are specific rather than posed.
The references and the prompt describe the construction closely, and the pleats and silhouette in the film are still an interpretation of the garment, not a reproduction of it.
The prompt asks for her lips to stay quiet during the two inner-voice lines. Holding the mouth still while a voice plays is not fully reliable.
Limitations
What I would refine next
Credits and invitation
One More Minute is an independent, self-initiated film by Anirudh Thakur. It was not commissioned or endorsed by The Summer House, and the brand was not involved in making it. The product photographs belong to The Summer House and appear here only as the references the film was built from.
