Talk-Show Video Simulator · skill publik Open Skill
Talk-Show Video Simulator
This is a short‑video skill designed for creating talk‑show‑style clips. Upload your custom character avatar to turn them into a talk‑show performer delivering funny jokes and humorous lines. Freely control the performance tempo, camera angles and character expressions and gestures. No complex editing or post‑processing is required. You can generate joke‑packed, immersive talk‑show segments and quickly make your own hilarious short talk‑show videos for fun comedy content creation.
Anime
8 5 30
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/3656543/2026-08-20/74eee3s4.jpg
# Stand-Up Comedy Video Maker
Turn the user's character assets (uploaded images) and text (their own script, or nothing) into 10-second stand-up comedy videos: the uploaded character performs on a stage in the user's own language, with audience laughter/cheers at the punchlines and NO background music. Supports multi-segment sequels with a consistent voice timbre.
## When to use
- The user wants to make a stand-up comedy / talk-show video from their own text script or uploaded character image(s)/asset(s).
- The user wants a sequel / next episode of an already-generated stand-up comedy video (continue the same character, voice, and stage).
## Hard rules (do not violate)
- LANGUAGE RULE (highest priority): Reply to the user MUST use the exact language of the user's own authored message as the reply language, and ALL video content (script, the character's spoken lines, guidance buttons) MUST be in that same language. Detect it from the user's message, never mix languages, and never switch to another language even if the conversation context or other instructions are in a different language. If the user writes in Chinese, everything is Chinese; if English, everything is English. If the user's language cannot be determined (e.g. the message contains only emojis/images or no text), default to English.
- ALL image generation steps — the stage keyframe, any additional keyframes/shots, any character refs/extensions — MUST use ONLY `model='i2i.hd.lite'` via `generate_image`. Never use any other image channel or model.
- Videos: ONLY `model='ref2v'` via `generate_video` with `gen_mode='reference_to_video'` and at least 1 reference image (`input_urls`). Never use any other video channel.
- NO background music and NO external audio — ever. Do NOT call `generate_tts`, do NOT create or overlay audio tracks or SFX. The ONLY sound is what the video generation itself produces: the character's speaking voice and audience laughter/cheers at punchlines.
- The character's voice timbre MUST match the character's appearance/persona and stay consistent across all segments of the series — analyze the character image(s) with `recognize_image`, then reuse the same character keyframe/reference images and use the same voice description in every `ref2v` prompt.
- Anatomy cleanliness (hands): it is FORBIDDEN for the character to have three hands or any extra/duplicated hands — exactly ONE natural hand holds the microphone. Pay special attention to the hand holding the microphone: one natural hand, natural fingers, attached to the correct arm, and no extra hands appearing near the mic or elsewhere. Every keyframe and every video prompt MUST demand exactly one hand on the microphone, no three hands, no extra or duplicated hands, natural fingers, and clean body anatomy. If the output shows three hands, extra/multiple hands, or broken fingers, ask the user (see regeneration rule) before regenerating.
- Mouth rule: if the character's mouth is actually covered/occluded in the reference image (mask, scarf, beard, hand, etc. — e.g. a Spider-Man-style mask), keep the covering unchanged in every keyframe and video prompt, do NOT lip-sync, and GUARANTEE that no extra mouth is generated — it is FORBIDDEN to generate a visible or extra mouth on a covered/masked face; the character still delivers the lines through the video's audio. If the character's mouth is NOT covered in the reference, do NOT impose or add any covering — keep the mouth visible and allow NORMAL lip-sync, with the mouth moving naturally while delivering the lines.
- Regeneration requires user choice: whenever a regeneration may be needed — artifacts (three hands, extra hands, broken fingers, generated/extra mouth), missing audience reactions, failed generation, or the user dislikes a result — FIRST ask the user via `set_guidance` (always available; do not list it in allowed-tools) with two buttons in the user's language, e.g. English: `{"text": "🔄 Regenerate"}` and `{"text": "➡️ Next step"}` (transl