Your Shota Can Chat with You · публичный навык Open Skill
Your Shota Can Chat with You
Your Shota Can Chat with You — публичный Open Skill на Digen. Откройте, чтобы увидеть, что создаёт этот процесс.
Anime
5 3 15
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/3947739/2026-08-21/7mx7gfvu.jpg
# Little Boy First-Person Chat Short Video Generator ## Hard Requirements (Read First) - All "must / forbidden / never / top-priority" rules are consolidated in the reference file `references/hard-requirements.md` - Before starting, you MUST call read_skill_file(name='boy-chat-video-maker', path='references/hard-requirements.md'), read it, and strictly follow every rule; the body steps below do not repeat them - Section 1 of the hard requirements is Language Consistency (Top Priority): user input language = generated language = communication language — this is a hard requirement, no exceptions - If a body step conflicts with the hard requirements, the hard requirements take precedence ## When to use - The user uploads one photo of a person and wants to create an 8-16 year-old little boy character, then make a chat-style vertical short video in first-person view (camera = user's eyes) - Story positioning: the boy cares and asks "what happened to you today", centered on the user's daily life — the boy asks / cares / follows up; the user has no lines (pauses for the user's response between the boy's lines) - The user wants to set the boy's personality (tsundere, soft and cute, energetic, aloof, etc.) for story-driven interaction; the interaction content is strictly clean and wholesome - Also use when redoing only one stage: character reference image, chat lines, scene storyboard, a single line's shot, BGM, final mix ## Post-generation self-check and confirmation (run at every generation stage) - Every generated image (portrait, storyboard image): call recognize_image to self-check - Every generated video (line shot, final cut): call recognize_video to self-check - Self-check content: whether the character looks like the original photo, whether the art style / temperament is consistent, whether the composition is compliant (above the calves), whether the scene lighting is consistent, whether the boy's dialogue box displays correctly (appears and disappears in order), whether the line is spoken correctly, whether the line language matches the user's input language, whether the language used to communicate with the user matches the user's input language (Section 1 hard requirement), whether the voice matches the chosen one (a boy uses a male voice), whether lip-sync is on time, whether the voice is clear, whether the camera is fixed without shaking, whether transitions are smooth - Honestly report the issues found in self-check to the user (no hiding, no embellishing), then use set_guidance to give two options "Regenerate" / "Continue to next step" and continue per the user's choice; never auto-redo, and never deliver an obviously flawed product as the final result ## Steps ### 1. Collect input and confirm requirements - Ask the user to upload 1 photo of a person; if not uploaded yet, request it first - Confirm with the user: - Age: 8-16 (default 12) - Art style: no conversion; follow the uploaded photo's art style (realistic / 2D, rendering style, realism) - Portrait temperament: default "cute and adorable" (soft and cute style); if the user specifies another temperament (handsome / aloof / cool, etc.), follow the user's request - Clothing: follow the photo, do not restyle on your own (if specific clothing is needed, ask the user to upload a corresponding photo) - Scene: follow the user's request (if not specified, confirm a suitable scene with the user before continuing) - Gender and voice: confirm the character's gender with the user (per the user's choice, usually matching the photo); set the voice by that gender — a boy character MUST use a boy voice (male voice), a girl character uses a girl voice (female voice); the voice can be a child's voice or an adult voice (adult voice allowed), chosen by the user; write the confirmed voice into all subsequent video prompts - Personality and chat topic are not confirmed yet; ask the user after the portrait is generated (see Step 3) - Use recognize_image to identify th