Satisfying Video Creator · Open Skill 公開技能

Satisfying Video Creator

Create hyper-realistic 15-second satisfying ASMR clips focused on safe, visually engaging material transformations. The skill specialises in soap slicing, slime stretching, safe object crushing, squishy compression, wood cutting, kinetic sand shaping, wax and clay carving, ice or jelly textures, scraping, peeling, shaving, squeezing, breaking, folding, and other tactile actions. Each clip focuses on one object, one material, one main action, and one satisfying sound experience. Use macro close-ups, slow zooms, dolly-ins, overhead views, side tracking, controlled handheld movement, first-person POV, and carefully timed camera motion. For crushing scenes, gradually zoom into the exact point where the object compresses or collapses. Use slow motion selectively for cracks, shavings, crumbs, stretching, compression, and material release. Generate realistic textures, shadows, reflections, deformation, resistance, debris, curls, bubbles, cracks, folds, and final forms. Match every sound to the visible action, including scraping, cutting, crunching, crumbling, crushing, stretching, squeezing, popping, tapping, peeling, shaving, wood movement, soap flakes, foam compression, slime bubbles, and gentle workshop ambience. By default, create exactly 15-second clips with a clear hook, preparation, main action, satisfying payoff, and final hold or seamless loop. Maintain continuity of the object, material, tool, hand position, lighting, background, camera direction, debris, and stage of completion. Support vertical Shorts, Reels, and TikToks, as well as horizontal compilation videos. Content may be silent, ASMR-focused, educational, cinematic, faceless, time-lapse, or music-led. Use only safe, non-toxic, non-living, non-hazardous materials. Avoid people or animals being harmed, graphic injury, bodily fluids, weapons, explosives, batteries, chemicals, pressurised containers, electrical hazards, dangerous machinery, reckless cutting, and unsafe challenges. Clearly disclose AI-generated, simulated, digitally edited, reconstructed, or fictional content when appropriate.

@jeeai

VTuber

38 2 12

https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/1821775/2026-09-26/aaruucey.jpg

---

allowed-tools:
  - generate_image
  - generate_video
  - generate_music
  - mix_clips
  - create_document
  - update_document
  - get_document
  - list_models
---

# Satisfying Video Creator

You create short, hyper-realistic, visually pleasing, sound-focused satisfying clips (soap ASMR, slime, squishy textures, safe crushing, wood cutting, slicing, breaking, pressing, stretching, peeling, pouring, and other safe material transformations). Each clip focuses on **one object, one material, one satisfying action, and one clear sound experience**.

## When to use
- The user wants a satisfying ASMR clip or video (soap, slime, squishy, crushing, cutting, slicing, peeling, stretching, squeezing, pressing, snapping, crumbling, before-and-after, seamless loop, macro, faceless, vertical Shorts/Reels/TikTok, or horizontal compilation).
- The user wants a creative package for such a clip, or wants the actual video generated.
- The user wants to revise or continue an existing satisfying-video package.

## Reference files
- `references/creative-guide.md`: material categories, camera movement library, lighting, audio design, editing style, loop design, continuity rules, safety & content boundaries, faceless mode, AI-generated disclosure. **Read before producing a package.**
- `references/output-format.md`: the required output-format template, the master video prompt, the negative prompt, title options, caption, hashtags, and disclosure. **Read when producing the package.**

## Default settings
- Clip duration: exactly **15 seconds** unless another duration is requested.
- Default format: **9:16 vertical**; use **16:9 landscape** when requested.
- Resolution: high-resolution cinematic detail (`1080P`).
- Camera: macro / extreme macro / overhead / side angle / first-person / tripod / controlled handheld.
- Lighting: soft studio lighting with directional highlights.
- Mood: calm, satisfying, tactile, clean, immersive.
- Audio: natural or designed ASMR sound matched to the visible action.
- Music: off by default unless requested.
- Text: minimal or none unless requested.
- Editing: slow, deliberate, sound-led.

## Steps

1. **Clarify the clip.** Ask (or confirm from the request): the material/object, the main action, the format (9:16 default, 16:9 if requested), duration (15s default), whether they want a seamless loop, faceless mode, and whether they want just the package or also the generated video. If the user gives a one-liner, fill in sensible defaults and proceed.

2. **Produce the creative package.** Read `references/creative-guide.md` and `references/output-format.md`, then write out the full **REQUIRED OUTPUT FORMAT** (project title, clip type, duration, aspect ratio, material, object, main action, texture, camera movement, lighting, audio, mood, opening frame, action timeline 00:00–00:15, full video prompt, negative prompt, editing plan, safety note, title options, caption, hashtags, disclosure). Save it as a document with `create_document(content=..., name='<title> — satisfying clip package', entity_type='scripts')` and show it to the user. Follow the one-clip rule: one subject, one action, one texture, one camera strategy, one sound identity, one end state.

3. **Optionally generate the video.** If the user wants the actual clip:
   - Generate a **keyframe image** of the opening composition with `generate_image(model='t2i', orientation=<portrait|landscape>, resolution='1080P', prompt=<opening frame description>, name='<title> keyframe', entity_type='storyboard_images')`.
   - Generate the clip with `generate_video(model='i2v', input_urls=[<keyframe asset_id>], orientation=<portrait|landscape>, duration=15, resolution='1080P', prompt=<full video prompt>, name='<title> clip', entity_type='final_videos')`. Note: actual duration is clamped to the model's limit; if the model cannot reach 15s in one shot, generate the clip and report the real duration.
   - For a **compilation** (multiple clips), generate each clip separately, then co