ASMR Cinematic Shorts Creator · skill pública Open Skill
ASMR Cinematic Shorts Creator
ASMR Cinematic Shorts Creator é uma Open Skill pública no Digen. Abra para ver o que este fluxo cria.
Anime
3 0 4
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4103102/2026-09-20/bw5jm3ha.png
# ASMR Cinematic Shorts Creator Turns an idea, a scene, a character, or a mix of uploaded image/character/scene references into a polished **horizontal→vertical cinematic YouTube Short**: fast cuts, no voiceover, only natural ASMR ambient audio, portrait 9:16, target duration ~140–170s. ## When to use - The user wants a vertical short/movie for YouTube from a fresh idea or uploaded image/character/scene references. - The user wants the signature house style: cinematic fast cut, no voiceover, natural ASMR ambient sound only, 140–170s. ## Memory (built-in) Preferences below are the defaults, but check the user's saved style first so you don't overwrite their choices. - Before starting, call `memory_query` (or `memory_whoami`) with a query about the user's short/movie style to load any saved defaults (aspect ratio, cut style, audio preference, target length, recurring mood/themes). - After a run, call `memory_store(kind='preference')` to remember any style/setting the user confirms (e.g. "prefers 9:16, no voiceover, natural ASMR ambient, cinematic fast cuts, ~150s"). Use the skill's name `asmr-shorts-creator` as the store `namespace`. ## Steps 1. **Clarify (1 round max).** Determine the concept/theme, what media the user is providing (fresh idea vs uploaded images/characters/scenes — treat each uploaded image as a shot seed), and the exact target duration (default ~150s, within the 140–170s band). Confirm the ASMR mood (rain, forest, fire, water, wind, city night, etc.). 2. **Write a fast-cut shot list.** `create_document(content=..., name="Shot list", entity_type="scripts", content_type="text/markdown")` with a numbered list of ~6s shots that together span the target duration. Keep it paced for fast cuts (each shot 5–8s, ~20–26 shots for 150s). 3. **Generate the scene images.** For each shot, `generate_image`. - If the user gave a reference image/character/scene, use `model='i2i'` and pass its `asset_id`/URL in `input_urls` to keep that look; otherwise use `model='t2i'`. - Always pass `orientation='portrait'`, `resolution='1080P'`, and `name` = the shot title, `entity_type='storyboard_images'`. Describe a cinematic, high-detail, fast-cut-ready composition in English, consistent across shots (same palette/lighting/mood). 4. **Generate the natural ASMR ambient track.** `generate_music` with `prompt` describing pure natural ASMR (e.g. "natural ASMR ambience only: gentle rainfall, distant thunder, rustling leaves, soft water lapping — no music, no beats, no voice"), `name="ASMR ambient", entity_type="music"`, and pass `duration` ≈ the target total length. No voiceover anywhere. 5. **Animate each scene to a clip.** For each shot image, `generate_video` with `model='i2v'`, `input_urls=[scene image asset_id]`, `orientation='portrait'`, `duration` ≈ that shot's length (5–8s), `resolution='720P'`. Prompt for subtle but alive cinematic motion that suits the fast-cut rhythm (e.g. slow tracking, drifting light, falling rain, gentle camera push) — keep cuts filmic, not static. 6. **Cut and mix.** `mix_clips` to concatenate all clips in shot order and add the ASMR music underneath: - `clips` = one entry with `videos` (all clip asset_ids in order), and `audio` = the ASMR track asset_id (set `delay=0`, `volume` ~0.8 so it sits as natural background). - Pass `orientation='portrait'`, `name="Cinematic ASMR short", entity_type='final_videos'`. If the ASMR track is shorter or longer than the mix, trim/loop it via `audio.start`/`duration` so it spans the whole length. 7. **Deliver.** Use `write_todos` to track steps 2–6 as you go, and `set_guidance` at the end to suggest next actions (e.g. restyle a shot, lengthen/shorten, add a different ASMR mood, tweak a scene). Reply in natural language with the final short's title and a short summary — don't paste long URL blocks. ## Notes - **No voiceover, ever** — audio is natural ASMR ambient only. If the user asks for narration, politely note this skill is ASMR-silent and offer to switch