Shared with Video Production Control : Storyboard Engineering、stage checks、material contract。
First clarify the audience's existing knowledge, the single learning objective, factual evidence, and the action they can take after watching. Source verification precedes expression; generated visuals are not treated as evidence of facts or real operations. Organize the opening question/concrete example/concept explanation/visible causality/conclusion, without applying a fixed business template.
Before designing the plan, first determine the output route with the user: browser final video or Jianying draft, and record output_target. The browser route can choose the seven storyboard categories; Jianying by default uses only ST-001–004 and checks the actual native effects, without automatically introducing JS or webpage prerendering. The two are respectively handled byBrowser ProductionandJianying Productionexecution, sharing this director workflow. Only when the user agrees to prerendering specific shots should exceptions be designed and recorded by the control, and auto is not treated as consent.
First the overall visual and motion direction, then shot by shot: id, purpose, frame range, voiceover anchor, ST type, visuals, action, material source and rationale, contact feedback, shot transitions. Use concrete metaphors to explain abstract concepts, and clarify the boundary between metaphor and fact. Character intention → action → environmental result → character feedback; actions are not merely decorative. The rhythm is establishing the situation → action explanation → result and handoff; duration follows the content and actual voiceover arrangement.
The shared directory contains only ST-001–ST-007: single image, video, background + transparent material, full-image limited frames, Remotion, Simo whiteboard, SRT whiteboard. Images, data, real screen recordings, and externally generated videos are selected according to purpose; the same knowledge topic can use different shot combinations. Text/highlights/charts/shots, etc. are arranged within the type, without creating additional top-level categories. AI can define new types for the current episode, and after verification the user decides whether to share them.
ST-001 single-image animation can choose full-screen mode or picture-book mode; picture-book mode is the arrangement of background, illustration area, and text area. Keyframes and environmental effects are optional, chosen according to content and actual implementation; these two modes are noted only in ST-001, without adding a new layout gallery or expanding other modes.
Read when the user chooses picture-book narrative to express knowledgePicture Book Production Guide. Each page uses one concrete event or causal step to explain the content, with a complete illustration centered and a short sentence below; factual information remains accurate, and the boundaries of the story metaphor are clear. This layout can be produced as either a browser final video or a Jianying draft, designing animation according to the chosen route, and not treating left-right chart explanation columns as the default picture book. Multiple pages are multiple ST-001 shots, without adding new categories; story short sentences and narration subtitles are designed separately.
The complete plan is submitted to the user for confirmation; after confirmation, provide video prompts and the required actual reference images, and collect all visuals. Only after all required materials pass technical and actual visual review can voiceover/animation/preview/compositing proceed. Record stages to resume subsequent work, and do not treat existing materials as a confirmed plan. When the SRT whiteboard has no subtitles yet, first review the line draft and data; during production, align using the actual voiceover subtitles, without estimating timing or rendering in advance.
Use shot by shotStoryboard Material Preparation Directorylist the corresponding documents, required materials, formats, and missing items. After confirmation, write specific and complete prompts for each missing material that can be generated and deliver actual reference images; when facts, operation records, originals, and voiceover timings are missing, provide a collection checklist. Generic prompts must not be used to replace the task for this shot, and TSX, drawing choreography, or post-dubbing subtitles must not be mistakenly listed as visual materials that must be provided before work begins.
Review check: whether the actions explain the concepts; whether the text preserves the facts; synchronization of contact and feedback; continuity of the subject's identity from beginning to end; whether subsequent whiteboard areas are not revealed in advance; whether screenshots/screen recordings are readable; whether the purpose of each shot is achieved. Technical verification is independent of user aesthetic acceptance, and no promise is made that the video will be popular or that the generation quality will pass.
The CapCut plan is delivered as an independent draft and actual material package, with shot comparisons, subtitle strategy, audio, and local font/effect conditions listed in the description. Passing the in-package check only proves that the materials, paths, and structure are complete; opening in the CapCut app and the actual export, as well as the dynamic aesthetics, are still subject to separate review, seeLocal Integration。