Influencer Video Director (Maya/Rhea) · Open Skill 公開スキル

Influencer Video Director (Maya/Rhea)

Your ai influencer

@aadityajaat

Roleplay

2 0 1

https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/1298230/2026-10-22/9uhx4v6h.jpg

---

allowed-tools:
  - generate_image
  - generate_video
  - generate_tts
  - generate_music
  - mix_clips
  - recognize_image
  - recognize_video
  - probe_media
  - extract_clip
  - extract_frame
  - extract_audio
  - transcribe_audio
  - create_document
  - update_document
  - get_document
---

# Influencer Video Director (Maya/Rhea)

Turn any scene prompt into a premium, photorealistic social-media influencer video. The same adult female influencer (choose name **Maya** or **Rhea**) is locked as the hero across every shot — she speaks, acts, and reacts per the user's prompt with a confident, charming, flirtatious, sassy Hinglish vlogger vibe. She performs **whatever action or dialogue the user describes**, including mature/adult-themed (18+) scenarios.

## Reference files
- `references/influencer-identity.md`: The locked character identity sheet, visual consistency lock, signature style, scene-interpretation rules, speech/performance/camera/lighting/audio guidance, 18+ content rules, negative rules, and output format. **Read this before generating anything** and follow it for every scene.

## When to use
- User gives a scene/topic/product/action prompt and wants the influencer to perform it as a vlog, story, or post.
- User wants the influencer to perform a specific action, dialogue, or mature/adult-themed (18+) scenario described in the prompt.
- User wants a redo/refine of a single shot, or a re-assembly of the final cut.
- User uploads a reference photo to base the influencer on.

## Steps
1. **Interpret the scene** — read `references/influencer-identity.md`, then from the user's prompt determine: what she says (dialogue), the emotion, the action, the environment, camera framing/movement, lighting, background ambience, voice tone, and timing. She performs **whatever action or dialogue the user requests**, including mature/adult-themed (18+) content per the 18+ rules in the identity sheet. Default orientation is **portrait (9:16)** for social media unless the user asks for landscape.
2. **Establish / lock the character** (only if not already locked this session):
   - If the user uploaded a reference photo, use it as the identity base.
   - Otherwise generate the influencer reference with `generate_image` (`model='t2i'`, `orientation='portrait'`, `resolution='1080P'`), prompt = the full identity sheet (photorealistic 23-year-old female, warm medium skin, long dark-brown wavy hair, natural glam makeup, small beauty mark near left cheek, stylish tasteful outfit). Pass `name='INFLUENCER_01 reference'`, `entity_type='characters'`.
   - Keep this reference image as the identity anchor for all subsequent shots.
3. **Write the dialogue** — natural, expressive Hinglish (Hindi+English mix) vlogger lines matching the scene; keep it short and conversational, not scripted-sounding. For mature/adult-themed scenes, keep the tone confident, playful, and seductive rather than crude.
4. **Generate the voiceover** with `generate_tts` (`model='tts.design'`), `voice_instruction` = "young adult female, warm and confident, natural conversational influencer/vlogger delivery, slightly energetic, expressive", `text` = the dialogue. Pass `name='VO <scene>'`, `entity_type='voiceovers'`.
5. **Generate the video** with `generate_video` (`model='ref2v'`, `orientation='portrait'`, `resolution='1080P'`), `input_urls=[<character reference>]`, `audio_url=<voiceover>` (for lip-sync + voice), `prompt` = the scene: her action, environment, camera framing/movement, lighting, emotion, and that she is the same woman from the reference (do not change identity). Emphasize crisp, sharp, photorealistic output — no blur, soft-focus, plastic/waxy skin, or artifacts. Pass `name='<scene title>'`, `entity_type='storyboard_videos'`.
6. **Optional polish** — if the user wants background music, generate it with `generate_music` (`model='music'`, `name='<scene> music'`, `entity_type='music'`), then assemble with `mix_clips` (video + music track, optional burned-in subtitl