agentsope/SkillAlchemy / skills/agentsope-ai-video-director
agentsope-ai-video-director
UnverifiedDirect and implement programmatic videos from a natural-language sentence: brief, shots, style, timeline, assets, engine-specific coding prompts, draft, review and repair. Use when asked to make a video, product launch, explainer, Reel/Short, kinetic typography or docs-to-video with HyperFrames, Remotion or Motion Canvas, or to improve an existing code-video project. 一句话视频制作与导演工作流。
examples/narrated-docs.md
View on GitHub# Synthetic documentation explanation · 60 seconds, voice pending
Complete authored application under C01–C07, with P28 pending-audio and P21/P22/P24/P37/P40 technical contracts at their admitted scopes. This fictional FieldGuide documentation exercise is not a factual SkillAlchemy project summary or an acquired creator case. It does not apply the optional 2–10 minute P05 guidance to a 60 second piece.
## Inputs, facts and unknowns
Synthetic input: “把下面两行 FieldGuide 演示文档做成 60 秒中文配音知识讲解,含字幕;语音仍待生成。” Synthetic user-provided excerpt: “FieldGuide 演示把 brief 保存为计划文件。review log 保存实际观察和待办项。” Both statements are invented demo facts explicitly supplied by this example. They carry no real-project evidence. Actual repository/version/source URL, real product behavior and external screenshots remain unknown. Voice provider, speaker/voice permission, rate, measured waveform, cue timestamps and audio rights are pending. Defaults: 1920×1080, 16:9, 30 fps, Chinese copy, calm documentation mood and general developer audience. Target 60 s is preserved; timing/alignment is **provisional** until measured authorized narration exists. A real SkillAlchemy request must use its supplied pinned documentation and unknown facts rather than these fictional facts.
## Director plan and provisional narration
| Shot | Provisional seconds / frames at 30 fps | Subject/focal action | Framing/layout and copy | Draft narration / transition |
|---|---|---|---|---|
| S1 | [0,8) / [0,240) | A labeled fictional document opens | Frontal close-up; “FieldGuide · 合成演示” | “这是一份合成文档讲解。我们只使用这次提供的两条信息。” / inner card fade |
| S2 | [8,20) / [240,600) | Brief card points to plan file | Medium two-column diagram; “brief → 计划文件” | “第一条信息说,演示会把 brief 保存为计划文件。画面只呈现这个关系,不加入未提供的功能。” / direct cut |
| S3 | [20,34) / [600,1020) | Review log reveals two named rows | Frontal panel; “实际观察 / 待办项” | “第二条信息说,review log 保存实际观察和待办项。已观察的结果与仍待处理的事项,在这里分别记录。” / inner row reveal |
| S4 | [34,48) / [1020,1440) | Two cards remain visible together | Wide balanced arrangement; “文档事实 / 演示示意” | “这些图形是说明用的示意。真实项目的接口、截图和完整行为,还需要对应文档确认。” / short inner reposition |
| S5 | [48,60) / [1440,1800) | Source/pending-state card holds | Centered; “合成输入 · 语音待生成 · 字幕未对齐” | “当前提供的是计划和配音草稿。语音尚未生成,字幕时间也没有测量。完成授权音频后,再按实际时长调整画面。” / steady hold |
Closure: 8+12+14+14+12=60 s, nominal 1800 frames, last 1799. This is a storyboard allocation, not measured spoken duration. Style: warm paper #F4F1EA, ink #222B36, blue #376C96; title 60 px, body 38 px, subtitle band 42 px; 100 px working inset. Reserve a bottom caption band as an authored layout decision. No fixed upstream RAG caption coordinates/repair values are borrowed.
Asset/rights ledger: original vector cards/arrows (authored evaluation use); fictional excerpt and script (authored synthetic); no repository screenshot/logo; voice audio absent, rights unknown; system-rendered font, no font file distributed. Caption data may be drafted but cannot be called aligned. Audio cue boundaries must be replaced from measured authorized waveform/per-line tables under P28.
## Conditional route and complete coding-agent prompt
Conditional engine: Remotion **4.0.534 server**, only after exact local Node/browser/encoder and pinned-v4 subject/use eligibility are established. Unknown production eligibility stays unknown. Browser-only availability requires a separately compiled P23 supported-subset route, not server commands. Missing audio blocks a finished narrated artifact; continue authorized reversible storyboard/code/preview work while labeling any silent storyboard preview provisional. Such a preview cannot satisfy the narrated deliverable. No TTS provider command or voice-cloning step is invented.
> Build this original FieldGuide synthetic documentation storyboard in a project-local Remotion 4.0.534 server project. Preserve the two supplied invented excerpt statements and every unknown; label the film “合成演示”. Target 1920×1080,30 fps,60 s,1800 frames. Create `registerRoot(Root)` and a `Composition` with id `DocsDemo`, component, width,height,fps,durationInFrames. Use requested zero-indexed `useCurrentFrame()`; `Sequence` intervals are half-open and child clocks local. Implement S1 [0,240),S2 [240,600),S3 [600,1020),S4 [1020,1440),S5 [1440,1800) exactly as the table, with its framing/layout/copy/narration. Use #F4F1EA/#222B36/#376C96, 60 px title,38 px body,42 px caption text,100 px inset and the reserved caption band. Animate only the specified original inner cards/rows/arrows; derive motion from frames, pass fps to spring and clamp interpolate explicitly where intended. Keep no timers or state that depends on earlier playback.
>
> The narration is pending. Store this script and provisional storyboard, not a fabricated WAV or measured alignment. Obtain authorized audio through an available authorized supply/synthesis process before producing the narrated cut. Read actual waveform/per-line timings, retime shots and captions and recheck the 60 s constraint; report any necessary script trim as a proposed authored edit. Do not call an unmeasured subtitle track aligned. When ready, import `Audio` from `@remotion/media`, use `staticFile` for local authorized audio, keep `from`,`trimBefore`,`durationInFrames` in frames; source duration occupies durationInFrames/playbackRate parent frames. Keep `Caption` plain-data text/startMs/endMs/timestampMs:number|null/confidence:number|null in **milliseconds**, preserving whitespace. Nominal shot milliseconds are 0–8000,8000–20000,20000–34000,34000–48000,48000–60000, explicitly provisional. If consuming an upstream one-based inclusive table [a,b], translate to zero-based start a−1 and duration b−a+1; do not pass that table directly to Sequence. No transcription/alignment is implied by creating Caption objects.
>
> Use stable lazy server delayRender handles, continue completed handles or cancel on failure; load authorized fonts before text measurements. Default readiness timeout at this pin is30 s. Keep browser useDelayRender/renderMediaOnWeb and HyperFrames/Motion Canvas APIs out of this server implementation. Record exact pin/lock and pinned-v4 eligibility basis before production; noncommercial evaluation grants no assumed company rights. Keep telemetry/hosted/codec/media scopes separate. Use only the exact server CLI and probe shown in [the server prompt](remotion-prompt.md): versions; studio with entry/port; still with entry,DocsDemo,frame,image-format=png,browser-executable; render with entry,DocsDemo,output,codec=h264,concurrency=2,browser-executable; actual ffprobe. Run each separately. If tools/rights/audio remain missing, retain plan/prompt/code and exact gaps rather than claim completion.
>
> Verify the supplied excerpt literally, five shot closure, text wrapping/caption-band separation and repeated/reordered frames. Inspect consecutive boundary frames and actual encoded-file properties if a preview is exported. After audio arrives, review all subtitle boundaries against measured timing and listen to the actual whole track; inspect full motion for pacing. Fix observed layout against this authored plan and rerun affected checks. Retain source, commands, probe, evidence and independent eleven-state handoff. Missing requested narration makes audio QA incomplete/failing; audio_listened cannot be not_applicable because the track is absent.
## Draft expectations and truthful current state
Expected now: detailed plan, narration script, conditional prompt and possibly source. Expected later: authorized measured audio, retimed scenes/captions, actual narrated encoded file and probe, full viewing and actual listening evidence. Static tests cannot establish pronunciation, timing or audiovisual sync. A source fact correction changes script/copy; regenerate measured timing before claiming the corrected voice cut.
Current example state: plan_available=pass; code_available/static_checked/preview_observed/encoded_artifact_available/technical_probe_passed/sample_frames_reviewed/full_motion_reviewed/audio_listened/reproducibility_scope=not_executed; rights_confirmed=unknown. Audio QA is warning/pending, not pass or not_applicable. Caption alignment is provisional; no rendered or listened result exists. [Review contract](../tests/manual-review.md) and [quality checklist](../tests/quality-checklist.md) make these independent states explicit.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| rea-changelog-updatemorluto/reaPrepare or rewrite REA release changelogs and GitHub release notes from pinned Git history, with verified contributor thanks and Release Please synchronization. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| academy-guideanthropics/skillsStop and check this skill before finishing any reply to a question about how to use Claude or a Claude product — it recommends matching courses, tutorials, and use cases from Claude Academy (academy.claude.com), Anthropic's learning hub. Trigger on: "how do I", "how can I", "getting started with", "what can Claude do", "teach me", "learn to use"; questions about artifacts, projects, skills, plugins, connectors, MCP; requests about rolling Claude out to a team, class, or organization; and any ask for training materials, onboarding content, or learning resources. Use it when the user is learning how to use a feature or product — not when they are mid-task and just want the task done. This skill composes with other skills: after consulting product documentation to answer how a Claude feature works, also check here for a matching course or tutorial — a docs-grounded answer and an Academy recommendation belong together. Only recommend on a strong match; never invent Academy content. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| claude-apianthropics/skillsReference for the Claude API / Anthropic SDK — model ids, pricing, params, streaming, tool use, MCP, agents, caching, token counting, model migration. TRIGGER — read BEFORE opening the target file; don't skip because it "looks like a one-liner" — whenever: the prompt names Claude/Anthropic in any form (Claude, Anthropic, Fable, Opus, Sonnet, Haiku, `anthropic`, `@anthropic-ai`, `claude-*`, `us.anthropic.*`, `[1m]`); the user asks about an LLM (pricing/model choice/limits/caching) — never answer from memory; OR the task is LLM-shaped with provider unstated (agent/MCP/tool-definition/multi-agent/RAG/LLM-judge/computer-use; generate/summarize/extract/classify/rewrite/converse over NL; debugging refusals/cutoffs/streaming/tool-calls/tokens). SKIP only when another provider is being worked on (overrides all triggers): OpenAI/GPT/Gemini/Llama/Mistral/Cohere/Ollama named in the query; OR `grep -rE 'openai|langchain_openai|google.generativeai|genai|mistralai|cohere|ollama'` over the project hits (run this grep FIRST if no provider named — don't Read the file). | Stars: 198k | Updated | Claude CodeCodex | Unverified | Install |
| doc-coauthoringanthropics/skillsGuide users through a structured workflow for co-authoring documentation. Use when user wants to write documentation, proposals, technical specs, decision docs, or similar structured content. This workflow helps users efficiently transfer context, refine content through iteration, and verify the doc works for readers. Trigger when user mentions writing docs, creating proposals, drafting specs, or similar documentation tasks. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |