Tentacle Crisis · skill pubblica Open Skill
Tentacle Crisis
Tentacle Crisis è una Open Skill pubblica su Digen. Aprila per vedere cosa crea questo workflow.
ShortDrama
18 4 13
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/3947739/2026-08-28/q7rd29b5.png
# Tentacle Play Vertical Video ## When to use - The user uploads **one character image** and wants an adult tentacle-play video: the character lies flat on their back in a **spread-eagle pose (arms and legs spread wide)**, tentacles wrap around the **arms, thighs and ankles**, output as a **9:16 vertical** video. ## Inputs - **Character image** (required): exactly one image of the character. - **Clothing image** (optional, max 1): an image of the outfit the user wants the character to wear. If provided, the character wears exactly that outfit; if not, the character keeps their original clothing. - **Tentacle color** (user-selectable): purple-black / pink. No default — ask if missing. ## Language - **Always use the user's input language** for every interaction: the tentacle-color question, button labels, narration, and the final delivery message. The generated output must stay consistent with the user's input language (fixed rule — see `references/hard-requirements.md`, rule 13). ## Hard Requirements All fixed rules — outfit (keep fully clothed), pose (spread-eagle), tentacle color, no clipping, style consistency, NSFW channels, silent video, identity, input quality, honest reporting, prohibited content, language — are defined in `references/hard-requirements.md`. **Read that file before generating** and follow every rule in it; they override any conflicting user request. ## Steps 1. **Read the character.** Call `recognize_image` on the uploaded image to note the character's key features (hair color/style, skin tone, face, body type, any distinctive marks) **and the art style** (anime / realistic / 3D render / painterly, line quality, color palette, rendering look) so identity AND style are preserved in the generated scene. If a clothing image was also uploaded, call `recognize_image` on it too and note the outfit's type, colors and style. 2. **Confirm the choices.** Check the user's message for the tentacle color. If it is missing, ask **one** short question **in the user's language** offering the options (purple-black / pink). Do not ask more than this one round. 3. **Generate the scene still** with `generate_image`: - `model='i2i'` (bare channel — NSFW-safe; do NOT use `i2i.hd`), `input_urls=[character image]` (plus the clothing image as the 2nd entry if one was uploaded), `orientation='portrait'`, `resolution='1080P'` - `name='tentacle-play scene still'`, `entity_type='storyboard_images'` - `prompt` = the **Scene still prompt** below, filled with the character features, the outfit (the uploaded clothing if provided, otherwise the character's original clothing), chosen tentacle color, and the art style noted in step 1. 4. **Offer the two buttons after the still.** In the same turn as showing the still, call `set_guidance` with exactly two actions: "Regenerate" and "Continue to next step" — **both translated into the user's input language** (e.g., Chinese: 重新生成 / 继续下一步; both default `send_message`). Then continue based on the user's choice: - **Regenerate** → redo step 3 (regenerate the still), then re-offer the two buttons. - **Continue to next step** → go to step 5. 5. **Animate the still** with `generate_video`: - `model='i2v.nsfw'`, `input_urls=[scene still from step 3]`, `orientation='portrait'`, `resolution='1080P'`, `duration=5`, `add_to_canvas=false` (intermediate — the silent version is shown instead) - `name='tentacle-play video'`, `entity_type='final_videos'` - `prompt` = the **Motion prompt** below, filled with the chosen tentacle color. 6. **Make the video silent.** Call `mix_clips` with a single clip whose video track is the video from step 5 and `volume=0.0` (mutes the original audio), producing the final silent video. `name='tentacle-play video (silent)'`, `entity_type='final_videos'`. 7. **Offer the two buttons after the video.** In the same turn as showing the silent video, call `set_guidance` with exactly two actions: "Regenerate" and "Done" — **both translated into the user's i