AI Influencer Creator · compétence Open Skill
AI Influencer Creator
One character. Every frame. The same face. Most AI portraits give you a different person each time. This skill builds ONE character — a sheet, a locked identity, a signature detail — and then keeps them identical across a photo set, a reel, a product shot, or a whole series. WHAT IT DOES • Generates a character reference sheet first: front and three-quarter, even light, plain background — the source of truth • Persists the bible in memory: face, body, wardrobe language, palette, and one unforgettable signature detail • Locks identity with four rules: always i2i from the approved frame, always describe the signature detail, one wardrobe language per set, one lens per set • Builds a six-frame photo set that actually proves consistency — front, three-quarter, environment, detail, motion, low angle • Shoots a reel in 3–5 second shots with one small motion each, because long shots are where faces drift • Optionally gives them a voice and keeps it, so they sound the same next time • Puts them to work: holding a product, appearing in a series, or starring in a short WHAT MAKES IT GOOD • Consistency is treated as the product — a pretty frame with a drifted face is a failure • One approved frame first, then chained from it for the rest of the set • Small motions only — no fast turns, no big expressions, no identity break • It remembers your character, so the next session opens with the same person GOOD FOR Virtual influencers, brand mascots, VTuber personas, consistent models for product shots, recurring casts for shorts, and fashion or beauty photo sets. NOT FOR Real people's likeness, public figures, minors in any sexualised framing, or adult content. Original characters only. Describe who they are and what they are for — it will build the sheet, lock the identity, and shoot the set.
Beauty
0 1 1
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4267104/2026-10-20/9m67auuq.png
--- allowed-tools: - generate_image - generate_video - generate_tts - generate_music - recognize_image - recognize_video - split_image - mix_clips - create_document - update_document - memory_whoami - memory_query - memory_store --- # AI Influencer Creator Build **one character** who stays the same face across every photo and every video, then put them to work: a photo set, a short reel, a product shot, a series. This skill's whole value is **consistency** — the same person, every time. ## When to use - The user wants a **virtual influencer, model, mascot or VTuber persona**. - The user needs the **same character across many images or videos**. - The user wants a **character sheet**, a **photo set**, or a **reel** for one character. - The user wants a character to **hold or wear a product**. - The user wants a **recurring cast** for a series of shorts. Do **not** use this for: a real person's likeness, minors in any sexualised framing, or impersonating a living public figure — see Notes. ## Steps ### 0. Intake — five decisions and nothing more Call `write_todos`: intake → character sheet → identity lock → photo set → reel → deliver. Ask at most **three** questions; choose the rest yourself and say so: 1. **Who are they?** — age range, vibe, one thing that makes them recognisable. 2. **What are they for?** — fashion, beauty, food, tech, lifestyle, gaming. 3. **Realistic or stylised?** — photoreal, semi-real, or anime. The other two decisions — **palette** and **signature detail** — you make. A character needs one unforgettable detail: a scar, a mole under the eye, silver streaks, a specific earring, a tattoo placement. ### 1. Character sheet — generate the bible first Before any scene, generate a **reference sheet** and treat it as the source of truth: ``` generate_image( model='t2i', prompt='character reference sheet, front view and three-quarter view of the same person: <age>, <hair colour and cut>, <eye colour>, <face shape>, <skin tone>, <body type>, <signature detail>, neutral expression, plain grey background, even studio light, photorealistic, no text', name='character sheet', entity_type='characters', orientation='portrait' ) ``` Then persevere it: ``` memory_store(key='character_<slug>', value='<name>; <age>; <hair>; <eyes>; <face shape>; <skin>; <body>; <signature detail>; <wardrobe language>; <palette>; <vibe>; sheet_asset=<ref>') ``` Show the sheet to the user and get a yes. **Everything downstream depends on this one frame.** ### 2. Identity lock — the rules that keep the face These rules are the skill. Break one and the character becomes someone else: - **Every image featuring the character uses `i2i`** against the sheet (or the best approved frame). Never `t2i` the face a second time. - **Describe the signature detail in every prompt.** If it is not in the prompt, the model drops it. - **One wardrobe language per set.** Same palette family, or it reads as a different person. - **Same lens per set** — 50mm portrait for close, 35mm for full body. Do not mix. - **Generate one approved frame first**, then chain from that frame for the rest of the set. ### 3. Photo set — six frames that prove consistency A usable influencer set is six frames, not sixty: | # | Frame | Purpose | |---|---|---| | 1 | 正面ポートレート | Face at 50mm, neutral background — the identity anchor | | 2 | 3/4アングル | Turned, same light — proves it is a 3D person | | 3 | 環境カット | In their world (street, café, studio), wider | | 4 | ディテール | Hands, jewellery, fabric — the signature detail up close | | 5 | 動きの中 | Walking, turning, hair moving — a frame from a video | | 6 | ローアングル | From below or above — proves the face holds from any angle | Generate them **from the approved frame**, one at a time, checking the face each time. ### 4. Reel — 3 to 6 shots, one motion each ``` generate_video( model='i2v', input_urls=[<approved frame>], prompt='<one motion: slow turn to camera / walking to