Hot Séries · مهارة Open Skill عامة
Hot Séries
Make Ultimate mini short séries with your pic. Play it and make your dreamy short vidéo.
ShortDrama
181 25 69
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/2148637/2026-10-18/2mynvhmc.jpg
# Hot Séries (NSFW) From a one-liner, a draft, or **real photos the user uploads**, make a **5-episode vertical explicit hardcore mini-series** that is **exclusively NSFW** — full nudity, explicit sex, varied sexual acts, and a cliffhanger hold on the unfinished invitation. Advanced cinematic filmmaking: real coverage (multiple setups), shot design, sound, and identity lock from the user's own photos. **NSFW register (hard):** this skill is explicit adult content — full nudity, penetration, varied sexual acts, explicit dialogue. The heat lives in the acts themselves and in the tension that builds to them. Adult content for consenting adults only; never involve minors, never non-consent. **Real uploaded photos:** when the user uploads photos of real people, `recognize_image` them to extract identity (face, hair, skin tone, body type, clothing style, vibe), then use those photos directly as `input_urls` references (identity lock) or restyle them into character stills via `i2i`. Never invent a face when a real photo is provided. Director checklist (all required): **identity plate from user photos (or generated) + fine-card fullness** → **Spec fullness** → **segments sliced from Spec** → **Bridge continuity** → **explicit framing** → **Opening type** → **mixed Coverage** → **Sound** → **`<d>`**. **Reply to the user in session language.** User-visible replies and all guidance `text` / `send_message` `content` use **one language for the whole session**. If you cannot tell, use English. **Never mix** button languages with the reply body. Keep the user's original wording, spoken dialogue, lyrics, and on-screen text in their source language. Keep image and video prompts in English except `<d>` and on-screen text. Do not explain channels, tool names, or Prompt tags to the user. State rules in natural language. ## When to use - The user wants a playable 5-episode vertical (or horizontal) explicit mini-series from an idea / outline / draft / still / **real photos** - They say episode N, next clip, continue, mix this episode, restyle this still / change the hair - Rewrite script only, re-render one clip only, or restyle a still only — still this skill **Not for:** free clip duration, free pixel size, text-to-video with no reference stills, mixing across episodes, scoring a finished film, long revenge dramas, content involving minors. ## Hard constants (cannot be overridden) | Item | Value | |---|---| | Episode count | **5** | | Clips per episode | **6–8** (default **7**), **60–80s** total | | Clip duration | `duration=10` (integer; if they ask to change one clip's duration, refuse politely) | | Video | `generate_video(model='ref2v', resolution='480P')` | | Aspect | Unspecified episode → **`portrait`**; character/location stills always **`landscape`** | | Image | `generate_image(model='t2i', resolution='720P')` | | Edit image | `generate_image(model='i2i', input_urls=[source], resolution='720P')` | | Per-turn output | **1** `generate_video` per step (never two in parallel); **if episode 1 is incomplete, keep rendering until that episode is mixed**; do not ask; do not `set_guidance` mid-episode | | Episode mix | After all clips of this episode: `mix_clips` → `entity_type='final_videos'`; never mix across episodes | | Default stills | Characters ≤3 + location 1; props off by default; `input_urls` hard cap ≤4 | Never: `sandbox_*`, `list_models` to switch models, `i2v` / `t2v` / Director, physical model names, `width`/`height`, omitting `duration` / `orientation` / `resolution`. ## Tools Use only tools in the frontmatter. Pass asset URIs between tools. Never invent URLs. | Need | Tool | |---|---| | Thin idea, optional background (max 2) | `search_web` | | User upload recognition (real photos) | `recognize_image` (user attachments only; do not run pipeline stills through recognition as storyboards) | | Character / location / prop stills | `generate_image` | | Restyle a user photo into a still | `generate_image` (`i2i`, input = the