bilibili-video-storyboard · git:20261015.2337206 · 2026-10-15 · sha256 989eb1d553175bab
bilibili-video-storyboard git:20261015.2337206A
Immutable. This exact content never changes and is served at /api/v1/blob/989eb1d553175bab.
--- name: "bilibili-video-storyboard" description: "Create a Bilibili video storyboard from a topic, title, outline, or script. This AI Bilibili storyboard maker turns a long-form video idea into a chapter-led shot list, Bilibili video script, camera direction, narration and B-roll cues, and one to four storyboard key frames for explainers, reviews, tutorials, vlogs, gameplay, animation, and creative videos." --- # Bilibili Video Storyboard Turn a Bilibili topic, title, outline, or script into a chapter-led video storyboard that an UP creator, director, camera operator, editor, or AI-video artist can review before production. Deliver a complete written shot list first, then create one to four selected key frames that make the most important visual beats easy to align. Use this Skill for Bilibili explainers, reviews, tutorials, Vlogs, gameplay, animation, creative videos, and channel-series planning. A request for one new cover belongs to `bilibili-thumbnail-maker`; a platform-neutral film, ad, or animation board belongs to `ai-storyboard-generator`; and approved plans can continue into `beatra-ai-video-studio` when the requested result is finished footage. Prove the sequence works before paying for key frames. Route publishing copy to `bilibili-publishing-pack`. ## Inputs and defaults Start with at least one of: a Bilibili topic, title, outline, script, chapter plan, or a brief that identifies the intended viewer and video outcome. Reuse known channel voice, series conventions, target duration, intended publishing view, product or factual must-keeps, people, locations, and visual references. For a source clip or Bilibili page, use a supplied transcript, outline, or exported screenshots as the planning source and label the role of every image reference. Default to a chapter-led creator plan with an opening hook, a clear section progression, one visual purpose per shot, a purposeful transition or retention reset between major sections, and a final viewer prompt. Keep the user's existing canvas and visual direction when known. Otherwise, defer the exact canvas choice until selected key frames are ready for paid work. ## Golden path The path is a topic or script, a beat sheet, human approval, then a limited set of key frames. 1. Turn the source into a free, reviewable Bilibili shot list. For every shot, show its chapter and story beat, time range, subject and action, frame size and composition, camera angle and intended movement, narration or sound cue, B-roll or graphic cue, and concise still-frame brief. 2. Review the complete order with the user. Select the one to four shots that need key frames, preserving the must-keeps, source-reference roles, canvas, and visual direction for each selected shot. Read [Bilibili storyboard planning and key frames](references/workflow.md) for the route-specific request shapes. 3. Before creating any key frame, read the live model card for the selected image capability. Show the route, complete prompt, reference order, canvas, count, model behavior, and current maximum price. Start paid work only after the user approves that frozen plan. 4. Treat each selected shot as its own `count: 1` paid image request. Give each one opaque, stable `client_request_id`; keep at most two image tasks in flight, track them to a terminal state, and then continue with any remaining approved shots. ## Execute once, review, and deliver Use only this Skill's bundled `scripts/mcp_client.py` for every remote Beatra operation. Do not configure or call a host Beatra Connector, and do not use REST/OpenAPI as a fallback. Upload local screenshots or reference images through the bundled client. Uploading provides an artifact reference; make visual claims only when the source or result is actually accessible to the host. Save each returned `task_id` and poll only that task through `beatra.tasks.get`. When a create response is genuinely unknown, replay only the byte-equivalent frozen request with the same `client_request_id`. When a task ID is missing, use `beatra.tasks.list`, confirm a candidate with `beatra.tasks.get` against the recorded request, and continue that task rather than submitting another paid request. Cancel only when the user asks; a `409` means the original task still runs and should keep being tracked. For accessible key frames, review the approved story beat, subject and action, composition, camera intent, canvas, and named must-keeps. Deliver the ordered shot list, the returned key-frame artifacts, actual dimensions and formats, resolved model, task identities, and returned charge facts. A visual correction or additional key frame is newly scoped paid work and needs its own approval. ## References by task - Read [Bilibili storyboard planning and key frames](references/workflow.md) when preparing a chapter plan, choosing an image route, creating a key frame, or recovering an image task. - Read [installation and authentication](references/installation-and-auth.md) and [installation registration](references/installation-registration.md) for authorization and the non-billable registration step. - Read [tasks and results](references/tasks-and-results.md), [billing, errors, and recovery](references/billing-errors-and-recovery.md), and [Bundled MCP Client diagnostics](references/mcp-connection.md) for shared runtime detail. - Updates and security: [updates and security](references/updates-and-security.md) - Read [uninstall and disconnect](references/uninstall-and-disconnect.md) when removing the Skill. ## Updates and security This Skill runs only the files installed with it and leaves them unchanged while it runs; it never downloads or executes other code. It talks to Beatra's official services over HTTPS, and a file you choose to upload goes only to the one-time upload address Beatra issues for it. Once a day at most, the bundled client reads Beatra's public version list. When the store or tool you installed from has a newer version, it prints one line with the command that updates it there. Updating stays your choice, through that store or tool. See [updates and security](references/updates-and-security.md).