AI Fashion Lookbook · Open Skill 公開スキル
AI Fashion Lookbook
Garment photos in. A finished lookbook out — 20 to 40 seconds, one consistent model, the clothes unchanged, fabric actually moving, and a cut that reads like a real campaign. This is not "a model wearing something similar". It keeps your garment exactly as it is and builds the shots a fashion buyer actually needs. WHAT IT DOES • Analyses the garment and locks silhouette, colour, pattern and construction as unchangeable • Casts one model for the collection and remembers them, so the next piece is shot in the same world • Builds a 10–14 shot lookbook: silhouette, fabric macro, motion, detail, fit, BACK, layering, close, environment, end card • Generates every frame with i2i against your garment photo, so the piece never changes • Animates the fabric — walk, turn, wind, a hand adjusting the collar — because movement sells the garment • Sets one light and one location per collection, which is what makes it read as a campaign rather than a slideshow • Delivers portrait 9:16, 4:5 feed or 16:9 site WHAT MAKES IT GOOD • The silhouette shot comes first — if a buyer cannot read the shape in 3 seconds, nothing else matters • At least two macro shots, because fabric texture is what justifies price • The back shot is mandatory, because it is what buyers always check • No shot where the face outranks the garment GOOD FOR Lookbooks, collection reels, seasonal drops, streetwear, luxury, minimal brands, outfit showcases, and vertical fashion reels for TikTok / Shorts / Reels / Instagram. NOT FOR Redesigning a garment, recreating another brand's campaign, or sexualised imagery. Original work only. Upload a garment photo and name the register — it will cast the model, shoot the lookbook and cut the reel.
Shopping
2 0 0
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4267104/2026-10-20/2r7dvzrf.png
--- allowed-tools: - generate_image - generate_video - generate_tts - generate_music - recognize_image - split_image - mix_clips - create_document - update_document - memory_whoami - memory_query - memory_store --- # AI Fashion Lookbook A garment photo in, a finished lookbook out — 20 to 40 seconds, one consistent model, the clothes unchanged, fabric actually moving, and a cut that reads like a real campaign. ## When to use - The user has **garment photos** and wants a lookbook video. - The user wants a **collection reel** or a **seasonal drop** film. - The user wants **one model wearing several pieces** with a consistent identity. - The user wants a **streetwear / luxury / minimal** brand video. - The user wants a vertical fashion reel for TikTok / Shorts / Reels / Instagram. Do **not** use this for: changing the garment's design, real brands' campaigns, or sexualised imagery — see Notes. ## Steps ### 0. Intake — the garment is the subject Call `write_todos`: intake → garment analysis → model + art direction → shot list → stills → motion → sound → cut → deliver. Ask at most **three** questions: 1. **What is being sold, and how many pieces?** — one hero piece, or a 3–5 piece capsule. 2. **What is the brand register?** — luxury, streetwear, minimal, soft, editorial. 3. **Where does it run?** — 9:16 for social, 4:5 for feed, 16:9 for site. If the user only uploads photos, infer the register from the garment and say so in one line. ### 1. Garment analysis — what may not change ``` recognize_image(image_url=<url>, prompt='Describe this garment for a fashion shoot: silhouette, fabric and weight, exact colour, construction details (seams, buttons, collar, hem), pattern and its scale, and how it would move. Do not invent materials or branding.') ``` Then persist so later pieces stay in the same world: ``` memory_store(key='fashion_brand_<slug>', value='<brand>; <register>; <palette: 2 + accent>; <model description>; <lighting language>; <surface/location>; <pieces shot>') ``` **Silhouette, colour and construction are locked.** You art-direct the light, the model, the movement — never the garment. ### 2. Model + art direction ``` memory_query(query='fashion lookbook model, brand register, previous pieces') ``` One model per collection, stored and reused: ``` memory_store(key='fashion_model_<slug>', value='<age range>; <height/build>; <hair>; <skin tone>; <signature look>; <neutral, non-expressive>') ``` Art direction decisions you make, and state in one line: - **Palette**: two neutrals + one accent. The garment owns the accent. - **Light**: one hard key for streetwear, one large soft key for luxury, window light for minimal. - **Location**: one surface, one environment. Changing both mid-reel breaks the collection. ### 3. Shot list — 10 to 14 shots for 30 seconds | # | Shot | Length | Function | |---|---|---|---| | 1 | シルエット | 3s | Full figure, wide, the whole garment readable | | 2 | 素材 | 3s | Macro: weave, seam, hardware | | 3 | 動き | 4s | Fabric in motion — walk, turn, wind | | 4 | ディテール | 3s | Collar, cuff, closure, print scale | | 5 | 着用感 | 3s | How it sits on the body; a hand adjusting | | 6 | 背面 | 3s | The back, which buyers always check | | 7 | レイヤー | 3s | Styled with a second piece | | 8 | クローズ | 3s | Chest-up, garment still dominant | | 9 | 環境 | 3s | Model small in the location, garment in context | | 10 | エンドカード | 3s | Brand / price / drop date, quiet | **Rules:** - **The silhouette shot comes first.** If the buyer cannot read the shape in 3 seconds, nothing else matters. - **At least two macro shots.** Fabric texture is what justifies price. - **The back shot is mandatory** for clothing. - **No shot where the face outranks the garment.** Faces sell people, not clothes. ### 4. Stills — garment first, model second ``` generate_image( model='i2i', input_urls=[<garment photo>], prompt='fashion editorial photograph, a model wearing <the exact garment>, unchanged silho