cclank/lanshu-create-ai-presenter-video

lanshu‑create‑ai‑presenter‑video

What it is – An Agent Skill that lets a coding‑assistant (Claude, Codex, Cursor, etc.) produce a complete explainer video from a short topic or script. The skill can follow two paths:

Route	Result	What you supply
Presenter	A digital‑human on‑camera, with live captions and keyword graphics	Topic or script + a portrait image you own (optional voice sample)
Styled explainer	Animated‑style video (no presenter) in one of 12 pre‑built visual or narrative styles	Topic or script (or your own narration)

The agent runs the repository’s scripts, creates a job folder, and steps through a gated workflow that records evidence on disk before moving to the next stage. All provider calls (voice synthesis, presenter video, lip‑sync, speech‑to‑text, etc.) are logged with model/parameter IDs, and cost‑guardrails ensure you see a billing statement before any paid request.

---

Key highlights

• Harness‑ and provider‑neutral – works with any Agent‑Skills client and any cloud or local model that can be invoked from the command line.
• Twelve ready‑made explainer styles – nine visual (editorial, HUD, notebook, pop‑up, comic, 3‑D one‑take, drafting sheet, chalkboard, clay town) and three narrative (derivation, case‑file, sitcom) kits, each with starter assets.
• Audio‑locked timeline – the final narration is the master clock; captions, presenter motion, cuts and graphics are all aligned to it.
• Evidence‑gated pipeline – eight production states (intake → verified) are computed automatically from the presence of artefacts, preventing accidental progression.
• Cost safety – pilot‑first generation, explicit billing preview, retry ceilings and task‑ID reuse to avoid duplicate charges.
• Verified delivery – master and share encodes are decoded and loudness‑checked before any file is published.

---

How it works (high‑level)

1. Init – init_job.py creates a self‑contained job directory and records inputs.
2. Pre‑flight – preflight.py validates the script, portrait rights, and other approvals.
3. Narration – Generates or accepts a voice track; word‑level timestamps are captured (MiniMax for styled route).
4. Presenter generation (presenter route) – Calls a video‑generation service to create a digital‑human clip, then lip‑syncs to the narration.
5. Audio‑driven edit – Captions, keyword graphics, opening/closing shots are assembled on the narration timeline.
6. Quality checks – qa.py / check_state.py verify decodeability, loudness (‑16 LUFS ±0.5), and visual consistency.
7. Render & delivery – finalize_delivery.sh produces master and share copies, a contact sheet and a delivery report.

The styled route follows a similar flow but skips presenter generation; instead each line of the script is turned into a short animated scene using the selected style kit.

---

Requirements

• Agent‑Skills harness (or any agent that can read files & run shell commands)
• Python 3.9+, FFmpeg (with ffprobe), Bash, jq, awk, sed
• Presenter route – access to a voice synthesis API, a presenter‑video generation API, and a lip‑sync tool (cloud CLI, API, or local model). Optional: HyperFrames for caption/graphics compositing.
• Styled route – Node.js (for HyperFrames), rsync, Python numpy, and a MiniMax API key for voice timestamps (or your own recorded narration).

---

Installation & quick start

# Clone into the skill directory of your harness (example for Claude Code)
git clone https://github.com/cclank/lanshu-create-ai-presenter-video.git \
  ~/.claude/skills/lanshu-create-ai-presenter-video

After cloning, either let the harness auto‑select the skill by describing the video you want, e.g.:

Turn this script and portrait into a 30‑second 9:16 presenter video with live captions.

Or invoke it explicitly:

SKILL_DIR=~/.claude/skills/lanshu-create-ai-presenter-video
python3 "$SKILL_DIR/scripts/init_job.py" \
  --job-dir ~/Videos/my‑presenter \
  --presenter-image ~/Pictures/presenter.png \
  --topic "Context engineering in one minute" \
  --duration 60 \
  --aspect 9:16 \
  --rights-confirmed \
  --adult-presenter-confirmed
python3 "$SKILL_DIR/scripts/preflight.py" ~/Videos/my‑presenter/job.json
python3 "$SKILL_DIR/scripts/check_state.py" ~/Videos/my‑presenter/job.json --write

For a styled explainer, replace --route styled and pick a style (e.g. v8‑chalkboard). The repository includes helper scripts (story.py, new_film.sh, qa.py, render.sh) that the agent runs automatically.

---

Testing

Two smoke‑test scripts are provided:

bash tests/smoke.sh               # presenter pipeline (no remote calls)
bash tests/explainer_smoke.sh     # styled pipeline (requires Node.js)

CI runs these on every push.

---

License

MIT – free to use, modify, and redistribute.

---

Bottom line – lanshu-create-ai-presenter-video is a ready‑to‑plug‑in skill that lets an LLM‑driven coding agent turn a brief idea into a polished explainer video, with strong safety, cost, and quality controls built into the workflow.