← Files PixVerseARCHIVED FILE
skills/pixverse-explainer/references/channels-and-formats.md
4.22 KB · Oct 2, 2026 · 00:28 UTC
# Channel, Format And Reference Routing Resolve the episode's question, audience, language, duration/aspect, storytelling channel, visual medium, audio mode, subtitles and cover separately. Use supplied answers; otherwise choose a coherent interpretation and state it briefly while proceeding. Do not restart an intake form for an already authorized production. Explicit duration, still/animation intent, reference look and sound instructions win over defaults below. | Channel | Story contract | Default when no look is supplied | | --- | --- | --- | | Explainer | Cold-open fact → escalating concrete evidence → a turn → a kicker; casual, deadpan, fast | Editorial Motion Graphics; offer Stickman or Whiteboard as the lighter alternative | | History | Sourced chronology told as a story about people doing things, witty and concrete | Editorial Motion Graphics; Paper Diorama for power/money/intrigue, Watercolor Chronicle for long-form, Mannequin for battles and expeditions | | Kids | Immediate plain-language question and answer; one physical example per beat, participating cast and a clear callback | Kids Studio 3D; four shots per 10s, Kids branch | | Picture Story | Continuous narration is the clock; dense edited images with consecutive-frame micro-variations | Papercraft or a requested medium; Picture Story branch | | Fairy Tale & Myth | A character's desire, obstacle, consequential choice and payoff; clearly frame mythology as mythology | Cinematic Storybook; storybook recipe | The channel does not constrain every subject: a myth-inspired visual treatment can explain airplanes without inventing magical lift, and a toy-like look can address adults without switching to children's language. A real reference takes precedence over a misleading catalog label. Inspect what it actually shows, then name that interpretation honestly. ## Mechanics And Audio - Fully animated narration: nominal 10s blocks, narration takes measured first, 3–5 shots per block cut to the spoken phrases (Kids four), proportional timing for a shorter last block. External voice owns speech; native video owns only agreed foley/ambience. - Explicit scene-based/still/frame-by-frame: Picture Story even on another channel. Use continuous voice first, not 10s voice fragments or slow zooms over a few pictures. - Kids talking cast: opt-in alternating narrator/native-dialogue blocks; preserve lip timing and at most two speakers in dialogue. Do not dub unrelated TTS over moving mouths. - Kids song: actual sung music first, measured musical sections, lyrics-led choreography. An instrumental bed, spoken rhyme or TTS cannot satisfy singing. Verify the current music route exposes lyrics/vocals before submission. - Long history: about 10 minutes or more loads the chapter/era-map branch before assets; it is a production-depth choice that works with several visual styles. Read `./branches.md` only for selected specialized mechanics. A request for only a script, prompt or storyboard ends there. A finished film request includes the actual final video. ## Reference And Acceptance Record Record each reference's role separately: topic/facts, style, identity, motion, or delivery format. A style donor is not evidence for dates, science or hardware. Website card prompts, advertised episode length and the actual playable excerpt may disagree; record all three and compare only the verified scope. Never claim to have reproduced an unseen full episode. Extract the visible medium, shape/line, palette, texture, stage depth, composition hierarchy and motion cadence. A blank intro frame does not establish the subject design; inspect a content-bearing frame. When a donor is supplied, bind its actual pixels to the style-key generation, then use the accepted key as the shared control. Do not merely mention its name. The style key shows the look on one real subject from the film. Check rendered blocks against the named defects in `./production.md`; wrong topic, wrong medium or missing requested speech fails even when every file decodes. Repairs target the failed layer/block, preserving accepted work. Re-preflight required generation under the user's existing authorization; no added confirmation gate. Keep empirically rendered coverage distinct from documented capability and untested modes.
SHA-256: 2dcd7760187538977da4adaa2af1b78cd5c8e959d724fc7d1857a01652ee9f60