ASMR Eating Video Creator · skill publik Open Skill

ASMR Eating Video Creator

ASMR Eating Video Creator adalah Open Skill publik di Digen. Buka untuk melihat apa yang dibuat alur ini.

@jeeai

VTuber

38 4 12

https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/1821775/2026-09-24/c6ethdrw.jpg

# ASMR Eating Video Creator

Create realistic, appetising, safe, and visually satisfying ASMR eating videos for YouTube, Shorts, Instagram, TikTok, and other platforms. Sound is the primary storytelling element; visuals are hyper-realistic, close-up, and texture-focused.

## When to use
- The user wants an ASMR eating / mukbang / food-tasting / sensory food video or package (crispy, crunchy, noodle/ramen, dessert/chocolate, fruit, street-food, traditional tasting, restaurant tasting, cooking-to-eating, hands-only, first-person, silent mukbang, gentle spoken ASMR, beverage pouring/sipping, packaging/unwrapping, cutting/slicing/stirring/dipping/pouring, texture-focused, or fictional/AI food experiences).
- The user wants to revise or continue an existing ASMR eating package, or generate only one part (food stills, scene visuals, voiceover, music, or the final video).

## Reference files
- `references/production-spec.md`: The full ASMR production spec (content types, food appearance, host appearance, audio priority, camera shots, lighting, table/environment, story structure, dialogue modes, host actions, food safety, reference handling, editing style). Read this before producing any package or asset.

## Steps
1. First use `write_todos` to list the steps for this request, and update the whole list as you progress.
2. Read `references/production-spec.md` and follow it for every decision (content type, camera, lighting, audio, safety, editing).
3. Clarify the request if needed: content type, food, duration (default 30–60s), aspect ratio (default 9:16 portrait unless another is requested), dialogue mode (silent / gentle spoken / food review), whether a reference person is shown or hands-only, and whether the user wants just the package or also generated assets.
4. Produce the creative package as a document via `create_document` (name like "ASMR Eating Package — <food/title>", `entity_type='scripts'`): video title, content type, duration, aspect ratio, camera plan, lighting plan, audio/sound plan, shot list, story structure, dialogue script (if spoken), food-safety notes, and publishing metadata (title/description/keywords/hashtags/thumbnail prompt). Keep the package concise and actionable.
5. Optionally generate assets (only if the user asks):
   - **Food / scene stills**: `generate_image` with `model='t2i'` (no reference) or `model='i2i'` (with a reference image). Use `orientation='portrait'` for 9:16, `landscape` for 16:9. `name` = shot/food title, `entity_type='storyboard_images'`. Prompts must describe realistic, appetising food per the spec (natural texture, accurate colour, realistic steam, believable moisture, correct shadows, physically accurate stretching/dripping/melting/breaking).
   - **Character / host stills** (if a person is shown): use `model='i2i'` with the authorised reference image to preserve identity; keep face, hair, skin tone, clothing, and proportions consistent. For hands-only, keep the same hands, sleeves, nails, utensils, table, and lighting throughout and do not show the face.
   - **Scene clips**: `generate_video` with `model='i2v'` from a still (or `model='ref2v'` with reference images to keep subject consistency; `model='ai2v'` when an uploaded audio track drives the scene). `orientation` must match the stills. `name` = clip title, `entity_type='storyboard_videos'`. Keep actions continuous — do not cut mid-bite/pour/slice/crunch.
   - **Voiceover** (gentle spoken / food review): `generate_tts` with `model='tts.design'` and a soft, calm, low-volume `voice_instruction`; or `model='tts.clone'` with `reference_audio_url` if the user wants a specific voice. `name` = line title, `entity_type='voiceovers'`.
   - **Music** (optional, kept below the ASMR triggers): `generate_music` with `model='music'`, `name` = track title, `entity_type='music'`.
   - **Final video**: stitch clips with `mix_clips` (`model='clip-mixer'`), layering the voiceover/music/ambient sound tracks under the video. `name` = final cut titl