Skip to content
Back to search

lanshu-create-ai-presenter-video

Turn a topic or finished script into a complete, publish-ready explainer video — led by an AI presenter from an authorized adult presenter image, or performed with no presenter in one of twelve explainer styles (nine visual styles and three narrative ones). Not for promos, ads, or footage montages. Use for new presenter or styled explainer videos and for continuing, revising, captioning, lip-sync repairing, or re-exporting an existing job. Use it whenever the user asks for 数字人口播 / 数字人讲解, for a 讲解视频 / 动画讲解 / 科普视频, asks which styles or templates are available (有哪些风格 / 看看风格 / 风格模板), or names one of the styles: 留白, 信号/科技感, 手帐/手绘, 立体书/纸艺, 波普漫画, 一镜到底/3D, 图纸与注脚, 黑板报, 黏土小城, 推演/动态图解, 案卷/调查复盘, 小剧场/生活场景. Keep model and provider selection capability-based and record the actual choices per job.

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/cclank-lanshu-create-ai-presenter-video/badge.svg)](/listings/cclank-lanshu-create-ai-presenter-video)

HTML

<a href="/listings/cclank-lanshu-create-ai-presenter-video"><img src="/api/listings/cclank-lanshu-create-ai-presenter-video/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

The repository directory is the skill. Clone it into your harness's skills directory and keep the folder name `lanshu-create-ai-presenter-video`, which the Agent Skills specification requires to match the skill `name`.

| Harness | Skills directory | Explicit invocation | |---|---|---| | Claude Code | `~/.claude/skills/` or project-level `.claude/skills/` | `/lanshu-create-ai-presenter-video` | | Codex | `~/.codex/skills/` | `$lanshu-create-ai-presenter-video` | | Other Agent Skills clients | See the client's documentation, linked from the [client list](https://agentskills.io/clients) | Client-specific |

For example, with Claude Code:

git clone https://github.com/cclank/lanshu-create-ai-presenter-video.git \
  ~/.claude/skills/lanshu-create-ai-presenter-video

Run `git pull` in the installed directory to update. If you use several harnesses, clone once into each skills directory.

Tagged versions and a ready-to-extract skill ZIP are available in [Releases](https://github.com/cclank/lanshu-create-ai-presenter-video/releases). See [CHANGELOG.md](CHANGELOG.md) for version history.

**Agents without Agent Skills support.** Clone the repository anywhere, then start the session with:

Read /path/to/lanshu-create-ai-presenter-video/SKILL.md and follow its workflow exactly.

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

SKILL.md · Manifest · 215 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

This file has changed since it was scanned, so this is not exactly the text that was scanned, and no line is marked. The findings are listed with the safety scan above.

  1. ---
  2. name: lanshu-create-ai-presenter-video
  3. description: >-
  4. Turn a topic or finished script into a complete, publish-ready explainer video — led by an AI presenter from an authorized adult presenter image, or performed with no presenter in one of twelve explainer styles (nine visual styles and three narrative ones). Not for promos, ads, or footage montages. Use for new presenter or styled explainer videos and for continuing, revising, captioning, lip-sync repairing, or re-exporting an existing job. Use it whenever the user asks for 数字人口播 / 数字人讲解, for a 讲解视频 / 动画讲解 / 科普视频, asks which styles or templates are available (有哪些风格 / 看看风格 / 风格模板), or names one of the styles: 留白, 信号/科技感, 手帐/手绘, 立体书/纸艺, 波普漫画, 一镜到底/3D, 图纸与注脚, 黑板报, 黏土小城, 推演/动态图解, 案卷/调查复盘, 小剧场/生活场景. Keep model and provider selection capability-based and record the actual choices per job.
  5. license: MIT
  6. compatibility: Works in any Agent Skills harness or shell-capable coding agent. Requires Python 3.9+, FFmpeg with ffprobe, bash, and jq; remote voice and presenter generation need network access and provider credentials. The styled explainer route also needs Node.js (npx) and rsync.
  7. ---
  8. # Lanshu Create AI Presenter Video
  9. Produce a verified explainer video from minimal inputs — led by a digital human, or performed in one of twelve explainer styles. The final narration is the master clock for presenter motion, captions, graphics, cuts, and delivery duration.
  10. ## Choose a route
  11. Every job takes one of two routes, recorded as `creative.route`. Infer it when the request is clear (cues below);
  12. otherwise ask once, with this table and the style gallery (`explainer/STYLES.md`; attach or link the images
  13. `explainer/gallery/nine-styles-kv-cache.jpg` and `explainer/gallery/three-narrative-styles-kv-cache.jpg`). Never fall back to `presenter` silently: it needs a portrait the user
  14. may not have.
  15. | Route | The video | The user provides | Paid generation | Format |
  16. |---|---|---|---|---|
  17. | `presenter` | A digital human presents; captions and keyword graphics support it | Topic or script + authorized presenter image | Voice + presenter video | Any; 9:16 by default |
  18. | `styled` | A performed explainer: every narrated phrase is acted out on screen in one of twelve styles (nine visual, three narrative), no presenter | Topic or script | Voice only | 16:9 |
  19. - A presenter image or the words 数字人 / presenter / 口播 / 真人出镜 point to `presenter`; 不要真人 / "no presenter" /
  20. "faceless", a named style, or 动画讲解 / 手绘 / 黑板 / 黏土 / 漫画 / 推导 / 复盘 / 小剧场 point to `styled`.
  21. - A vertical platform (抖音 / 视频号 / 小红书 / 竖屏 / 9:16) is a format request, not a route: `styled` is 16:9 only.
  22. If it comes with no portrait or with styled cues, ask (a 16:9 styled film vs a 9:16 presenter video) instead of
  23. switching routes.
  24. - Narration the user already recorded is the locked audio on either route. Never pass it as `--voice-sample` (that
  25. flag is for voice cloning). On `styled` use `story.py story --audio <file>` and ask for the matching script text
  26. (or transcribe it and have the user check it).
  27. - To re-style an earlier job, keep its script and locked audio (no new voice cost), build a new film from the same
  28. `story.json` in the new style, and re-approve the storyboard. Ask which job when it is not obvious.
  29. - This skill makes explainers. For 宣传片 / 广告 / 发布大片 / footage edits, say so, and offer a product explainer only
  30. if the user wants one.
  31. - For `styled` also choose `creative.explainer_style` (`v1-editorial` … `v12-sitcom`). Recommend two or three from the
  32. topic and audience (one reason each) using `explainer/STYLES.md`; when the user is unsure, build drafts in the nine
  33. visual styles (no extra paid calls once the audio is locked, or with `--dry`) and let them pick from a still grid.
  34. Settle the style before the script is locked: each style shapes how the narration is written and voiced
  35. (`explainer/STYLES.md`, "按风格写稿"). The three narrative styles (`v10-reasoning` 推演, `v11-casefile` 案卷,
  36. `v12-sitcom` 小剧场) go further: use the style's suggested arc when it helps the explanation, adapting chapters and
  37. order to the topic. The examples are starting points; keep room for other storytelling choices. To compare
  38. styles first, use `--dry` drafts of a first script, then adapt it to the chosen style.
  39. - `styled` follows [styled-explainer.md](references/styled-explainer.md) and reuses this skill's job directory, state
  40. machine, approvals, and delivery checks. The styles are designed to carry the explanation on their own; if a user
  41. also wants their digital human in a styled film, that file describes the method (time the film to the presenter's
  42. locked narration, give the presenter its own space) as an optional addition, not a separate route.
  43. How to ask, when the request leaves the choice open — one message, in the user's language:
  44. 1. The two routes, one line each: what the video looks like, what the user must provide, what is paid, and the
  45. format (presenter: any ratio, 9:16 by default; styled: 16:9 only). Recommend one for this topic and say why (e.g. "no portrait yet, or a concept that needs to be shown → styled"; "personal brand,
  46. talking to camera → presenter").
  47. 2. For `styled`, assume the user does not know the styles yet: show the whole menu — the two gallery images plus the
  48. twelve one-line entries of "风格菜单" in `explainer/STYLES.md` — and mark the two or three that fit this topic and
  49. audience, one reason each. Offer "build drafts in the nine visual styles and pick from a still grid" as the
  50. alternative. Do the same, without the route table, when the user only asks which styles there are.
  51. 3. Proceed with the recommendation if the user agrees or does not mind; record the choice in `job.json`.
  52. ## Required outcome
  53. - Start from a topic or script and, for the presenter route, one authorized image containing one clear adult presenter.
  54. - Deliver a fully decoded master video, a smaller share copy, captions, production records, and separate machine and visual QA notes.
  55. - Keep the workflow portable across providers, models, aspect ratios, languages, and durations.
  56. ## Operating rules
  57. - Confirm image rights, adult status, remote-upload approval, and voice-cloning authorization before the relevant remote action.
  58. - Treat remote generation as potentially billable. Before the first paid call, state the uploaded assets, requested seconds or units, known cost, pilot size, retry ceiling, and expected output. Reuse an approval already given for that exact plan.
  59. - Never infer a real voice from an image. Use an authorized sample or record a selected stock voice.
  60. - Lock the complete narration before presenter generation, caption timing, or final scene boundaries.
  61. - Prefer one presenter image, one voice identity, one visual treatment, and one continuous presenter source.
  62. - Mute video sources in the final composition. Route only the approved external narration and intentional mix tracks.
  63. - Preserve provider request bodies and task IDs without credentials or expiring URLs. Poll interrupted work before considering resubmission.
  64. - Use numeric checks for technical faults and normal-speed visual review for identity, mouth timing, blinking, gestures, hands, lighting, and continuity.
  65. - Stop after three rejected paid candidates, or before a change that materially affects cost, privacy, voice, appearance, or provider.
  66. - Do not claim completion until the final files fully decode, the contact sheet or full playback has been reviewed, and `check_state.py` reports `verified`.
  67. ## Start or resume a job
  68. For a new job, read [generation.md](references/generation.md). Set `SKILL_DIR` to the directory that contains this file; its location depends on the harness. Then run:
  69. ```bash
  70. SKILL_DIR=/path/to/lanshu-create-ai-presenter-video
  71. python3 "$SKILL_DIR/scripts/init_job.py" \
  72. --job-dir ~/Videos/my-presenter-video \
  73. --presenter-image ~/Pictures/presenter.png \
  74. --topic "用一分钟讲清楚上下文工程"
  75. ```
  76. For the styled route add `--route styled --explainer-style v3-notebook` (no presenter image needed).
  77. Use `--script` for an existing script file. Optional flags include `--voice-sample`, `--supporting-media`, `--duration`, `--aspect`, `--width`, `--height`, `--fps`, `--watermark`, and `--cta`. The initializer copies every input into the job and records job-relative paths, so the job directory is self-contained; keep later artifact paths job-relative too.
  78. Inspect the actual source image and listen to any voice sample. Record the manual review and approvals in `job.json`, then run:
  79. ```bash
  80. python3 "$SKILL_DIR/scripts/preflight.py" ~/Videos/my-presenter-video/job.json
  81. ```
  82. For an existing job, read `job.json`, current artifacts, task IDs, and QA reports, then run `check_state.py` (below) to find the earliest unfinished state. Resume there without regenerating accepted work.
  83. ## Production state machine
  84. Advance a job only when its evidence exists:
  85. ```text
  86. intake
  87. → content_locked passing preflight report, script, beat sheet
  88. → audio_locked decodable final audio, ASR report
  89. → visual_plan_locked timeline, storyboard, plan.status "approved"
  90. → presenter_generated recorded main_presenter capability, selected video, visual review
  91. → composition_checked composition report
  92. → rendered decodable render with video and audio
  93. → verified master, share, delivery report with passing output loudness
  94. ```
  95. On the `styled` route there is no presenter: `presenter_generated` is reached by the performed film instead
  96. (`artifacts.story`, `artifacts.film_project`, and the visual review).
  97. Never edit `state` by hand. Record artifacts in `job.json`, then let the checker compute the state:
  98. ```bash
  99. python3 "$SKILL_DIR/scripts/check_state.py" ~/Videos/my-presenter-video/job.json --write
  100. ```
  101. Without `--write` it only reports. It exits non-zero when the recorded state claims more than the evidence supports, and lists what the next state is missing.
  102. ### 1. Lock content and audio
  103. Read [generation.md](references/generation.md).
  104. 1. Turn a topic into one spoken content spine, or polish a supplied script without changing factual meaning silently.
  105. 2. Save the production script, beat sheet, pronunciations, and narration sections.
  106. 3. Generate the complete approved narration with one voice configuration.
  107. 4. Normalize sections consistently, run ASR on the final audio, and correct material omissions, additions, numbers, names, or repeated speech.
  108. 5. Record real durations. These durations now define the timeline.
  109. ### 2. Plan and generate the presenter
  110. Choose a presenter-led, screen-demo, or mixed-explainer route. Use supporting media only when it proves or clarifies a spoken point.
  111. When the provider caps request or reference-audio duration below the narration length, plan the split from the locked audio's ASR timings instead of guessing:
  112. ```bash
  113. python3 "$SKILL_DIR/scripts/plan_segments.py" \
  114. --timings ~/Videos/my-presenter-video/qa/asr/sentences.json \
  115. --audio ~/Videos/my-presenter-video/assets/audio/final/narration.wav \
  116. --cap 15 --whole-seconds \
  117. --output ~/Videos/my-presenter-video/qa/requests/segment-plan.json
  118. ```
  119. Save the plan path in `plan.segment_plan`. Generate every segment from the same image, seed, framing, light, and motion constraints.
  120. Generate a short, low-cost pilot before a full run. Prioritize identity and mouth timing for the main track. Use one controlled gesture for an actionful opening or close. When body motion is accepted but mouth timing fails, preserve the motion plate and apply a dedicated lip-sync repair with the locked audio.
  121. Archive the prompt, parameters, provider/model/version, region, task ID, requested seconds, and acceptance notes.
  122. ### 3. Edit the video
  123. Read [editing.md](references/editing.md).
  124. Build a deterministic timeline driven by the locked audio. Keep authored start, duration, and source offset independent. Add captions and keyword callouts only after audio and selected media are final.
  125. When using HyperFrames, load the current `hyperframes`, `general-video`, and relevant domain instructions. Run the project check, inspect transition and emphasis frames, open the Studio preview, and wait for final visual approval before rendering.
  126. ### 4. Verify and deliver
  127. Read [qa-recovery.md](references/qa-recovery.md).
  128. Render at delivery quality, watch the complete video, and finalize it:
  129. ```bash
  130. bash "$SKILL_DIR/scripts/finalize_delivery.sh" \
  131. ~/Videos/my-presenter-video/renders/rendered.mp4 \
  132. ~/Videos/my-presenter-video/outputs \
  133. my-video
  134. ```
  135. The finalizer preserves the input aspect ratio, performs two-pass program loudness normalization, creates master/share encodes, fully decodes them, measures their delivered loudness, counts black and frozen-picture events, and produces a nine-frame contact sheet. It builds everything in a temporary directory and publishes nothing unless every check passes. The target defaults to `-16 LUFS`; set `PROGRAM_LUFS` (for example `PROGRAM_LUFS=-14`) when the destination requires another.
  136. Inspect the contact sheet, review any freeze events against intentional still shots, record the outputs and delivery report in `job.json`, and run `check_state.py --write` to reach `verified`.
  137. ## Default behavior for minimal input
  138. - Infer language from the request.
  139. - On the presenter route, use 9:16, 1080×1920, 30fps unless the intended platform suggests another format. The styled route is always 16:9, 1920×1080, 30fps.
  140. - Preserve a supplied script's natural duration; for a topic, target 45–75 seconds.
  141. - Use a suitable stock voice when no authorized voice sample exists.
  142. - On the presenter route, use a presenter-led layout with a designed hook, 2–4 useful beats, and a concise close; on the styled route, the style's starter sets the layout.
  143. - Omit music and promotional CTA unless requested or clearly justified.
  144. - Keep styling credible, contemporary, readable, and safe for the destination platform.
  145. ## Required job artifacts
  146. ```text
  147. job.json
  148. docs/{SCRIPT,BEAT_SHEET,TIMELINE,STORYBOARD}.md
  149. assets/source/
  150. assets/audio/{reference,raw,final}/
  151. assets/video/{candidates,selected,render}/
  152. assets/captions/
  153. qa/{requests,asr,contacts,reports}/
  154. renders/
  155. outputs/
  156. ```
  157. ## Reference routing
  158. - Read [styled-explainer.md](references/styled-explainer.md) whenever `creative.route` is `styled`; it maps every state above onto the explainer tools in `explainer/`. Choose styles with `explainer/STYLES.md`.
  159. - Read [generation.md](references/generation.md) for intake, content, voice, tool selection, presenter prompts, paid generation, or provider changes.
  160. - Read [editing.md](references/editing.md) for timeline construction, screen-demo layouts, openings, closes, captions, keyword callouts, previews, and exports.
  161. - Read [qa-recovery.md](references/qa-recovery.md) before accepting media or delivery, and whenever lip sync, identity, hands, exposure, freezes, captions, audio, or remote jobs fail.
README.md · README · 339 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

This file has changed since it was scanned, so this is not exactly the text that was scanned, and no line is marked. The findings are listed with the safety scan above.

  1. # lanshu-create-ai-presenter-video
  2. **English** | [简体中文](README.zh-CN.md)
  3. [![Agent Skills](https://img.shields.io/badge/Agent_Skills-SKILL.md-111827)](https://agentskills.io)
  4. ![Harness Neutral](https://img.shields.io/badge/Harness-Neutral-8B5CF6)
  5. ![Provider Neutral](https://img.shields.io/badge/Provider-Neutral-0EA5E9)
  6. ![Routes: Presenter | Styled](https://img.shields.io/badge/Routes-Presenter_%7C_Styled-0EA5E9)
  7. [![Explainer Styles: 12](https://img.shields.io/badge/Explainer_Styles-12-E8590C)](explainer/STYLES.md)
  8. [![Rendered with HyperFrames](https://img.shields.io/badge/Rendered_with-HyperFrames-2563EB)](https://hyperframes.heygen.com)
  9. <br>
  10. [![Validate Skill](https://github.com/cclank/lanshu-create-ai-presenter-video/actions/workflows/validate.yml/badge.svg)](https://github.com/cclank/lanshu-create-ai-presenter-video/actions/workflows/validate.yml)
  11. [![License: MIT](https://img.shields.io/badge/License-MIT-F4C430.svg)](LICENSE)
  12. [![Release](https://img.shields.io/github/v/release/cclank/lanshu-create-ai-presenter-video)](https://github.com/cclank/lanshu-create-ai-presenter-video/releases/latest)
  13. [![Python 3.9+](https://img.shields.io/badge/Python-3.9%2B-3776AB?logo=python&logoColor=white)](https://www.python.org/)
  14. [![FFmpeg Required](https://img.shields.io/badge/FFmpeg-Required-007808?logo=ffmpeg&logoColor=white)](https://ffmpeg.org/)
  15. [![Node.js for the styled route](https://img.shields.io/badge/Node.js-styled_route-5FA04E?logo=nodedotjs&logoColor=white)](https://nodejs.org/)
  16. [![GitHub stars](https://img.shields.io/github/stars/cclank/lanshu-create-ai-presenter-video?style=flat)](https://github.com/cclank/lanshu-create-ai-presenter-video/stargazers)
  17. <br>
  18. [![X @LufzzLiz](https://img.shields.io/badge/X-@LufzzLiz-000000?logo=x&logoColor=white)](https://x.com/LufzzLiz)
  19. [![Author: 岚叔 Lanshu](https://img.shields.io/badge/Author-%E5%B2%9A%E5%8F%94_Lanshu-C2410C)](https://x.com/LufzzLiz)
  20. > Turn a topic or script into a verified, publish-ready explainer video — presented by a digital human from one authorized portrait, or acted out on screen in one of twelve explainer styles — driven by the coding agent you already use.
  21. `lanshu-create-ai-presenter-video` is an [Agent Skill](https://agentskills.io) that takes an AI agent through the full production of an explainer video. On the **presenter route** that means script, narration, presenter generation, lip-sync, captions and keyword motion graphics, editing, rendering, and quality assurance. On the **styled route** the narration is performed scene by scene in one of twelve styles (nine visual, three narrative), from a word-timed story to a rendered film. Every stage is gated by evidence on disk, and nothing is delivered until the output passes decode and loudness checks.
  22. ## Highlights
  23. - **Harness-neutral.** A standard `SKILL.md` that loads in Claude Code, Codex, Gemini CLI, Cursor, OpenCode, GitHub Copilot, and other Agent Skills clients. Any agent that can read files and run shell commands can follow it directly.
  24. - **Provider-neutral.** On the presenter route, voice, presenter video, lip-sync, and speech recognition are selected by capability at run time; each job records the provider, model, parameters, and task IDs it actually used. The styled route voices with MiniMax (word timestamps come from the voice engine) or takes narration you voiced or recorded elsewhere.
  25. - **Twelve performed explainer styles.** Nine visual styles — editorial, tech HUD, notebook, paper pop-up, pop comic, 3D one-take, drafting sheet, chalkboard, and clay town — plus three narrative styles: derivation, casefile, and life-scene sitcom. Each has a reusable kit, starter, and example project.
  26. - **Audio-locked timeline.** The approved narration is the master clock for presenter motion, captions, cuts, and final duration, which keeps lip-sync and scene boundaries aligned.
  27. - **Evidence-gated workflow.** Eight production states, from `intake` to `verified`, computed from artifacts instead of declared by hand.
  28. - **Cost guardrails.** Pilot-first generation, an explicit billing statement before the first paid call, retry ceilings, and task-ID recovery to avoid duplicate charges.
  29. - **Verified delivery.** Master and share encodes are fully decoded and loudness-checked before any file is published.
  30. ## Two routes
  31. | Route | The video | You provide | Paid generation | Format |
  32. |---|---|---|---|---|
  33. | **Presenter** | A digital human presents; captions and keyword graphics support it | Topic or script + authorized portrait | Voice + presenter video | Any; 9:16 by default |
  34. | **Styled explainer** | No presenter: every narrated phrase is acted out on screen in one of twelve styles (nine visual, three narrative) | Topic or script (or your own recorded narration) | Voice only (none with your own narration) | 16:9 |
  35. The agent infers the route from your request or asks once, with the style gallery. All styles share one contract —
  36. word-timed captions, chapter chips, a performed closing line, a one-frame recap. Nine are visual styles that differ in
  37. how ideas are acted out: editorial paper, tech HUD, notebook and pen, paper pop-up book, pop comic, 3D one-take,
  38. drafting sheet, chalkboard, and a clay town. Below is the same moment of the bundled KV cache explainer in all nine:
  39. ![The same moment of the KV cache explainer in nine styles](explainer/gallery/nine-styles-kv-cache.jpg)
  40. Three are narrative styles, where the look comes with a way of telling and the script can draw on its suggested arc: **推演**
  41. (derive it on one canvas, formulas and geometry bound by colour), **案卷** (investigate it on an evidence board, from the
  42. anomaly to the closed case) and **小剧场** (a life-scene sitcom: a misconception, a bill or a queue that shows the cause,
  43. one sentence that maps the analogy to the real term). Below is the same KV cache topic told the three ways:
  44. ![The KV cache topic told in the three narrative styles](explainer/gallery/three-narrative-styles-kv-cache.jpg)
  45. A new topic becomes a playable draft in any style in a few commands — narration, word timing, captions, chapters,
  46. closing line, recap, and sound are generated; the agent then performs each line's scene. See
  47. [`explainer/STYLES.md`](explainer/STYLES.md) to choose a style and
  48. [`references/styled-explainer.md`](references/styled-explainer.md) for the workflow — including an optional method for
  49. bringing your digital human into a styled film.
  50. ## What you provide
  51. | Input | Required | Notes |
  52. |---|---|---|
  53. | Topic or finished script | Yes | A topic becomes a 45–75 second script; a supplied script keeps its natural length. |
  54. | Presenter image | Presenter routes | One clear adult presenter, with confirmed rights to use the image. Not needed for the styled explainer. |
  55. | Voice sample | No | Used only with explicit cloning authorization; otherwise a stock voice is selected and recorded. |
  56. | Supporting media | No | Screen recordings, images, B-roll, or brand assets, used where they prove or clarify a spoken point. |
  57. | Delivery preferences | No | Platform, duration, aspect ratio, style, watermark, music, and call to action. |
  58. ## How it works
  59. Presenter route:
  60. ```text
  61. Topic or script + authorized portrait
  62. │
  63. ▼
  64. Lock script and full narration ──► narration becomes the master clock
  65. │
  66. ▼
  67. Low-cost presenter pilot
  68. │
  69. ▼
  70. Presenter generation (split at real pauses when a provider caps duration)
  71. │
  72. ▼
  73. Audio-driven edit: captions, keyword graphics, cover, close
  74. │
  75. ▼
  76. Technical and visual QA
  77. │
  78. ▼
  79. Master, share copy, contact sheet, and delivery report
  80. ```
  81. Styled route:
  82. ```text
  83. Topic or script ──► choose a style (menu + galleries, or drafts in the nine visual styles)
  84. │
  85. ▼
  86. Script written for the style ──► voice with word timestamps (or your own narration)
  87. │
  88. ▼
  89. Draft film from the style's starter: captions, chapters, closing line, recap, sound
  90. │
  91. ▼
  92. Storyboard ──► every line performed on screen, on its words
  93. │
  94. ▼
  95. QA (check, stillness, cover) ──► render ──► master, share copy, subtitles, delivery report
  96. ```
  97. Each state requires evidence before a job may advance. `check_state.py` computes the state from the job's artifacts:
  98. | State | Evidence required |
  99. |---|---|
  100. | `intake` | Job created |
  101. | `content_locked` | Passing preflight report, script, beat sheet |
  102. | `audio_locked` | Decodable final narration, ASR report |
  103. | `visual_plan_locked` | Timeline, storyboard, approved plan |
  104. | `presenter_generated` | Recorded presenter capability, selected video, visual review |
  105. | `composition_checked` | Composition report |
  106. | `rendered` | Decodable render with video and audio |
  107. | `verified` | Master, share copy, and a delivery report with passing output loudness |
  108. On the styled route, `audio_locked` also needs a voiced `story.json` (a `--dry` draft never counts), and
  109. `presenter_generated` is reached by the performed film and its visual review instead of a presenter video.
  110. ## Requirements
  111. - An Agent Skills harness, or any coding agent that can read files and run shell commands.
  112. - Python 3.9+, FFmpeg with `ffprobe`, Bash, `jq`, `awk`, and `sed`.
  113. - For the presenter route: access to at least one voice synthesis, presenter video generation, and lip-sync capability — a cloud CLI, an API, or a local model — and optionally a deterministic timeline compositor such as HyperFrames for captions, motion graphics, and final rendering.
  114. - For the styled explainer: Node.js (HyperFrames runs with `npx`), `rsync`, Python with `numpy` (sound bed; point
  115. `PYTHON` at it), and a MiniMax API key (`MINIMAX_API_KEY`) for the voice and its word timestamps — or your own
  116. recorded narration.
  117. ## Installation
  118. The repository directory is the skill. Clone it into your harness's skills directory and keep the folder name `lanshu-create-ai-presenter-video`, which the Agent Skills specification requires to match the skill `name`.
  119. | Harness | Skills directory | Explicit invocation |
  120. |---|---|---|
  121. | Claude Code | `~/.claude/skills/` or project-level `.claude/skills/` | `/lanshu-create-ai-presenter-video` |
  122. | Codex | `~/.codex/skills/` | `$lanshu-create-ai-presenter-video` |
  123. | Other Agent Skills clients | See the client's documentation, linked from the [client list](https://agentskills.io/clients) | Client-specific |
  124. For example, with Claude Code:
  125. ```bash
  126. git clone https://github.com/cclank/lanshu-create-ai-presenter-video.git \
  127. ~/.claude/skills/lanshu-create-ai-presenter-video
  128. ```
  129. Run `git pull` in the installed directory to update. If you use several harnesses, clone once into each skills directory.
  130. Tagged versions and a ready-to-extract skill ZIP are available in [Releases](https://github.com/cclank/lanshu-create-ai-presenter-video/releases). See [CHANGELOG.md](CHANGELOG.md) for version history.
  131. **Agents without Agent Skills support.** Clone the repository anywhere, then start the session with:
  132. ```text
  133. Read /path/to/lanshu-create-ai-presenter-video/SKILL.md and follow its workflow exactly.
  134. ```
  135. ## Quick start
  136. Most harnesses select the skill automatically from its description, so you can describe the video you want:
  137. ```text
  138. Turn this script and portrait into a 30-second 16:9 presenter video with live captions.
  139. ```
  140. For a styled explainer you don't need to know the styles yet — ask, and the agent shows the twelve with pictures and
  141. recommends a few for your topic:
  142. ```text
  143. Make a 40-second explainer about RAG, no presenter. Which styles are there?
  144. ```
  145. Use the explicit invocation from the table above to guarantee selection. The skill works in the language of your request.
  146. The agent runs the bundled scripts itself. To set up a job manually, point `SKILL_DIR` at your installation:
  147. ```bash
  148. SKILL_DIR=~/.claude/skills/lanshu-create-ai-presenter-video
  149. python3 "$SKILL_DIR/scripts/init_job.py" \
  150. --job-dir ~/Videos/my-presenter-video \
  151. --presenter-image ~/Pictures/presenter.png \
  152. --topic "Context engineering in one minute" \
  153. --duration 60 \
  154. --aspect 9:16 \
  155. --rights-confirmed \
  156. --adult-presenter-confirmed
  157. ```
  158. Complete the manual review and upload approvals in `job.json`, then run preflight:
  159. ```bash
  160. python3 "$SKILL_DIR/scripts/preflight.py" ~/Videos/my-presenter-video/job.json
  161. ```
  162. After each stage, record its artifacts in `job.json` and let the checker compute the state:
  163. ```bash
  164. python3 "$SKILL_DIR/scripts/check_state.py" ~/Videos/my-presenter-video/job.json --write
  165. ```
  166. A styled explainer by hand, from a job directory (`X=$SKILL_DIR/explainer`):
  167. ```bash
  168. python3 "$SKILL_DIR/scripts/init_job.py" --job-dir . --route styled --explainer-style v8-chalkboard --topic "RAG in 40 seconds"
  169. python3 "$X/tools/story.py" story # story/script.md → voice + word times → story/story.json
  170. bash "$X/tools/new_film.sh" v8-chalkboard story film # a playable draft; perform each line in film/beats.js
  171. python3 "$X/tools/qa.py" film qa/film # check, stillness, snapshots, cover
  172. bash "$X/tools/render.sh" film rag outputs # master, share copy, subtitles
  173. ```
  174. ## Scripts
  175. | Script | Purpose |
  176. |---|---|
  177. | `init_job.py` | Creates a self-contained job directory, copies all inputs into it, and records job-relative paths. |
  178. | `preflight.py` | Validates inputs, manual review, and approvals; separates local errors from remote-generation blockers. |
  179. | `plan_segments.py` | Splits the locked narration at real ASR pauses when a provider caps request or reference-audio duration. |
  180. | `check_state.py` | Computes the evidence-backed production state and exits non-zero when the recorded state overclaims. |
  181. | `finalize_delivery.sh` | Builds master and share encodes, verifies full decode, delivered loudness, and black or frozen frames, then publishes them with a contact sheet and report. |
  182. ## Explainer tools
  183. Under `explainer/tools/`, used on the styled route:
  184. | Tool | Purpose |
  185. |---|---|
  186. | `story.py` | `script.md` → MiniMax voice per line with the engine's word timestamps → `story.json`; `--audio` for your own narration, `--dry` for a silent draft, `--check` for the plan and estimated length. |
  187. | `new_film.sh` / `new_topic.sh` | A playable film from a style's starter (or drafts in the nine visual styles at once). |
  188. | `docs.py` | Script, beat sheet, timeline, and a storyboard table from the story. |
  189. | `qa.py` | `hyperframes check`, snapshots at every chapter, stillness pairs, a cover frame, and a composition report. |
  190. | `render.sh` | Render, finalize with `finalize_delivery.sh`, write subtitles (SRT); `DRAFT=1` makes a 720p review copy for a phone. |
  191. | `srt.py`, `still_grid.py`, `fonts.py`, `sync.sh`, `stamp.py` | Subtitles, a 3×3 style comparison still, font subsets, and project plumbing. |
  192. ## Reference guides
  193. The agent loads each guide only when it reaches the matching stage, which keeps context usage low.
  194. | Guide | Covers |
  195. |---|---|
  196. | [`generation.md`](references/generation.md) | Intake, script and narration, capability selection, billing gates, segment planning, presenter prompts |
  197. | [`editing.md`](references/editing.md) | Timeline contract, segment seams, openings and closes, captions, keyword graphics, preview and export |
  198. | [`qa-recovery.md`](references/qa-recovery.md) | Acceptance gates and recovery playbooks for lip-sync, identity, seams, loudness, and remote tasks |
  199. | [`styled-explainer.md`](references/styled-explainer.md) | The styled explainer route: story pipeline, style starters, performing scenes, and an optional method for adding a digital human |
  200. | [`explainer/STYLES.md`](explainer/STYLES.md) | The style menu and gallery, common words → styles, and how to write the narration for each style |
  201. | [`explainer/KITS.md`](explainer/KITS.md) | The kit contract, the aesthetic baseline, the starters, and the tools |
  202. | [`explainer/GOTCHAS.md`](explainer/GOTCHAS.md) | HyperFrames pitfalls met while building the styles |
  203. ## Repository layout
  204. ```text
  205. lanshu-create-ai-presenter-video/
  206. ├── SKILL.md # Skill entry point: metadata and workflow
  207. ├── README.md
  208. ├── README.zh-CN.md
  209. ├── agents/
  210. │ └── openai.yaml # Codex display metadata; other harnesses ignore it
  211. ├── assets/
  212. │ └── job.template.json # Job manifest template
  213. ├── references/ # Stage guides loaded on demand
  214. ├── scripts/ # Job setup, gates, segment planning, delivery
  215. ├── explainer/ # Styled explainer engine: core runtime, twelve style kits + starters, tools, examples
  216. │ ├── STYLES.md # Style chooser with the gallery
  217. │ ├── KITS.md # The kit contract and tools
  218. │ ├── kits/<style>/ # One look per style, and starter/ = the film minus the topic
  219. │ ├── stories/ # Example scripts (KV cache) and a draft test story
  220. │ ├── examples/kv-cache/ # The finished KV cache film in each style
  221. │ └── tools/ # story.py, new_film.sh, qa.py, render.sh, srt.py, …
  222. └── tests/
  223. ├── smoke.sh # End-to-end test with synthetic media
  224. └── explainer_smoke.sh # Styled route: intake, story, starter sync, state
  225. ```
  226. ## Defaults
  227. | Setting | Default |
  228. |---|---|
  229. | Format | 9:16, 1080×1920, 30 fps (styled explainer: 16:9, 1920×1080, 30 fps) |
  230. | Duration | 45–75 seconds from a topic; natural length for a supplied script |
  231. | Voice | Stock voice unless an authorized sample is provided; styled route: MiniMax `Chinese (Mandarin)_Reliable_Executive`, speed set per style |
  232. | Structure | Hook, 2–4 content beats, concise close |
  233. | Music and call to action | Off unless requested |
  234. | Loudness | −16 LUFS ±0.5 LU, true peak ≤ −1 dBTP; override with `PROGRAM_LUFS` |
  235. ## Safety and cost controls
  236. - Image rights and adult-presenter status are confirmed before any upload.
  237. - Voice cloning requires explicit authorization, and a voice is never inferred from an image.
  238. - Before the first paid call, the agent states the uploads, requested seconds, known price, pilot size, and retry ceiling.
  239. - When a provider caps duration, the quote uses the total requested seconds from the segment plan.
  240. - Interrupted remote tasks are polled by their saved task ID before any resubmission.
  241. - Work stops after three rejected paid candidates, with a summary of the recurring failure and remaining options.
  242. - On the styled route the only paid call is the voice — one call per line, cached, so an edit re-voices only that line; drafts in the nine visual styles cost nothing more.
  243. ## Privacy
  244. - The repository contains no API keys, access tokens, signed URLs, or user media.
  245. - Provider request records are stored without credentials or expiring URLs.
  246. - Preflight and delivery reports store file names or job-relative paths only, never absolute machine paths.
  247. - The styled explainer downloads font subsets from Google Fonts and jsDelivr (only the characters used); word timing
  248. comes from the voice provider's own timestamps, so no speech model is installed; snapshots are taken with
  249. `--describe false`, so frames stay local.
  250. ## Testing
  251. The smoke test drives a synthetic job from `intake` to `verified` without calling any remote service. It requires FFmpeg, `jq`, and Python 3.9+:
  252. ```bash
  253. bash tests/smoke.sh
  254. bash tests/explainer_smoke.sh # styled explainer; also needs Node.js and rsync
  255. ```
  256. CI runs metadata validation, a portability check, syntax checks for the explainer engine, and both smoke tests on every push and pull request.
  257. ## Contributing
  258. Issues and pull requests are welcome. Reports from different harnesses and providers are especially valuable, as are improvements to the workflow, compatibility, and quality checks. To add another explainer style, follow the kit contract in [`explainer/KITS.md`](explainer/KITS.md) and the starter brief in [`explainer/STARTER-BRIEF.md`](explainer/STARTER-BRIEF.md).
  259. ## Author
  260. Created and maintained by **岚叔 (Lanshu)**. Follow on X for updates and new AI video workflows:
  261. [![Follow on X](https://img.shields.io/badge/Follow_on_X-@LufzzLiz-000000?style=for-the-badge&logo=x&logoColor=white)](https://x.com/LufzzLiz)
  262. ## License
  263. Released under the [MIT License](LICENSE). Fonts, libraries, and services the explainer engine uses are listed in [`explainer/THIRD_PARTY.md`](explainer/THIRD_PARTY.md).

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.