Vincentwei1021/video-talkcraft

Agent skill that turns Claude Code / Codex into a motion-design studio for voiceover-driven explainer videos — word-level voiceover sync, 109 motion recipe cards, an anti-slideshow camera system, Remotion rendering.

What it solves

video-talkcraft transforms a script and a voiceover recording into a high-quality, motion-graphics-heavy talking-head video. It eliminates the manual effort of aligning text to audio, designing complex animations, and ensuring visual layout safety for educational, product review, or news commentary videos.

How it works

It functions as an AI agent skill (designed for Claude Code or Codex) that orchestrates a production pipeline:

  1. Audio Alignment: Uses ASR backends (FireRedASR2-CTC or faster-whisper) to generate character-level timestamps for the script.
  2. Shot Design: The agent analyzes the script to create a "SHOTBOOK" (storyboard) based on a library of 78 motion recipe cards.
  3. Remotion Implementation: The agent writes React-based video code using the Remotion framework to render dynamic text cards, B-roll, and camera movements.
  4. Quality Assurance: A triple-verification system checks for static frames, audio energy levels, and temporal defects using automated scripts.

Who it’s for

Content creators, educators, and reviewers who produce talking-head or explainer videos and want professional-grade motion graphics without manual editing in traditional software.

Highlights

  • Character-Level Sync: High-precision audio-to-text alignment (median deviation of 20–40ms).
  • 78 Motion Recipe Cards: A comprehensive library of self-contained Remotion tsx source code for various visual effects.
  • Anti-PPT System: A seven-layer model ensuring continuous camera curves and parallax planes to prevent static frames.
  • Automated Layout Safety: Uses OpenCV YuNet to detect face bounding boxes and ensure elements don't obstruct the speaker's face.
  • Triple Verification: Automated checks for static frames, sound effect cues, and frame-group reviews to ensure professional polish.

Related

  • Project
    thu-ml/Causal-Forcing

    A framework for high-quality real-time interactive video generation using Causal ODE or Causal Consistency Distillation to enable efficient few-step autoregressive diffusion.

  • Project
    Thmen/EGVSR

    A PyTorch implementation of EGVSR, an efficient video super-resolution framework that uses subpixel convolution to speed up inference for high-quality video upscaling.

  • Project
    gamedev-skills/awesome-gamedev-agent-skills

    A library of 73 AI‑agent “skills” (engine‑specific code guides, discipline, genre and workflow templates) plus a router that auto‑detects your game engine and loads only the needed pieces, enabling agents like Claude, Cursor, Gemini‑CLI, etc., to generate game code on demand.

  • Project
    huangserva/ComfyUI_MiniMaxH3_Director

    A collection of ComfyUI workflows for MiniMax H3 Director, enabling text-to-video, image-to-video, and advanced video-to-video editing with reference materials.