Digen Mini-Series Maker (Drama & Sensual) · Open Skill 公開スキル
Digen Mini-Series Maker (Drama & Sensual)
Digen Mini-Series Maker — это профессиональная ИИ-режиссура для создания полноценных драматических мини-сериалов. Ключевые особенности и функции: - Трёхъязычный замок (Tri-Lingual Lock): Навык автоматически распознает русский, английский или китайский язык пользователя и мгновенно переключает весь интерфейс. - Свобода в одежде + Фиксация лиц: Революционная функция Hard Face Token Lock намертво удерживает уникальные черты лица, расу и пропорции героев, но позволяет менять им гардероб в каждой серии под нужды сценария. - Качественная драматическая озвучка: В пайплайн встроен инструмент generate_tts, создающий глубокие, эмоциональные голоса для мужских и женских диалогов. - Живая микромимика: В RM35-видеопромпты зашиты алгоритмы естественного дыхания, моргания и плавных движений камеры (Anamorphic Cinema). - Скетч-контроль 4K: Сначала вы проверяете статичные кадры на сходство и анатомию, и только после вашего одобрения запускается видеогенерация. - Gapless Audio Mix: Сквозное микширование звука, исключающее заикания или прерывания музыки на склейках серий.
ShortDrama
52 8 26
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/3447002/2026-09-18/yazwkdvf.png
# Digen Mini-Series Maker
Engine for multi-episode dramatic mini-series with strict face/persona matching, user-defined wardrobe (never changed by AI), TTS voiceover, and a trilingual runtime (RU / EN / ZH).
## When to use
- The user wants to produce a dramatic (optionally sensual) mini-series episode from character/location reference images plus plot context.
- The user wants to continue to the next episode, redo a sketch, or re-render a clip of an existing episode.
## Reference files
- `references/prompt-templates.md`: English prompt templates for the photo sketches and the video clips. Read it before generating any sketch or clip.
## 1. Trilingual global protocol (read once)
- Detect `user_lang` from the user's first authored text: contains Cyrillic → `user_lang=ru`; contains Han characters → `user_lang=zh-CN`; otherwise/latin → `user_lang=en`.
- Every user-facing interaction (UI cards, canvas names, buttons, logs, subtitles) MUST match `user_lang`.
- TECHNICAL LAYER: all video prompts, prompt blocks, asset tracking and JSON matrices MUST always stay in English for platform execution.
## 2. Input capture (image core)
Ask the user to provide the core assets for the current episode:
1. 👩 **Image 1 (female identity sample):** face structure, eyes, skin tone, bone structure, AND her exact outfit (this is the user-defined clothing — preserved as-is).
2. 👨 **Image 2 (male identity sample):** masculine facial features, gaze, build, bone structure, AND his exact outfit (user-defined — preserved as-is).
3. 📸 **Image 3 (universal location):** set design, architecture, lighting references for the scene.
4. 📄 **Plot context:** character dialogue lines, a scene script, OR a short situation description (one-liner) — see 2b.
Mode buttons (localized to `user_lang`): [📱 Vertical 9:16] · [📺 Horizontal 16:9] · [🔥 Enable light erotic] · [🎬 Pure Drama only].
Locked variables: `selected_aspect` (portrait/landscape), `erotic_modifier` ('true'/'false'), `user_lang`. The chosen format is STRICTLY fixed for the whole episode: 9:16 → `orientation='portrait'`, 16:9 → `orientation='landscape'`. Apply this same `orientation` to EVERY asset in the pipeline (sketches, video clips, final cut) — never mix formats.
## 2b. Automatic 60-second plot from a short situation description
If the user gives only a short situation description (no full script or dialogue), auto-write a complete 60-second episode script from it:
1. Assign character names by gender: the female lead gets a female name, the male lead gets a male name (in `user_lang`).
2. Expand the one-liner into a full 60-second mini-episode arc: establish the scene, the conflict/tension, and the emotional payoff, timed to ~60s.
3. Split it into clean **female-lead** and **male-lead** dialogue lines (dialogue text in `user_lang`; technical prompt layer stays English).
4. Present the generated 60-second script to the user for approval and call `set_guidance` with localized buttons (RU: [✅ Сюжет готов, продолжай] · [✏️ Переписать сюжет]; EN: [✅ Plot approved, continue] · [✏️ Rewrite plot]; ZH: [✅ 剧情完成，继续] · [✏️ 重写剧情]).
5. Use the approved dialogue lines for TTS in section 3 and for audio-driving the clips in section 5.
## 3. Dramatic voiceover pipeline (generate_tts)
Before shooting, split the script lines into voice segments with `generate_tts` (`model='tts.design'`, `voice_instruction` in English). Keep the female and male VO segments separate — they will drive each video clip's lip sync in section 5:
1. **Voice 1 (female lead):** `generate_tts(text=<female_lines>, model='tts.design', voice_instruction="Mature, warm, dramatic female voice, cinematic delivery.", name='Female Lead VO', entity_type='voiceovers')`.
2. **Voice 2 (male lead):** `generate_tts(text=<male_lines>, model='tts.design', voice_instruction="Deep, husky, cinematic baritone male voice.", name='Male Lead VO', entity_type='voiceovers')`.
3. Combine the two voice tracks into one continuous `episode_audio_track` with `mix_clip