motion-studio
by mercadoa1234-arcDANTE
v3 · The ultimate teaching-studio skill. Generate complete narrated educational, scientific, and engineering videos that combine: parametric CAD (build123d) with exploded views and orbital cameras (pyvista), Manim motion graphics and equation reveals, source-document image insertion (PDF page screenshots, Substack post screenshots, photographs, figures), composite overlays, AI-driven 3D reconstruction handoffs, and high-fidelity browser-GPU renders — all narrated with bundled neural Kokoro-82M CPU TTS (seven voices, no network). Audio-first pipeline: phrase-aware Kokoro chunking gives natural rhythm to short phrase rhythms ("never three, never six, never nine") and paragraph-level breath to long sections. Every shot picks its render engine (pyvista · manim · composite · image · bom · title) and the orchestrator dispatches. Output: soft-sub MP4 (mov_text track + sidecar SRT, no pixel burn-in) with -14 LUFS audio mastering, gentle denoise, optional subtle reverb. Trigger phrases: "make a teaching video", "explain X with animation and narration", "narrate this paper", "adapt this Substack post into a video", "3blue1brown style explainer", "CAD walkthrough with formulas", "design and animate a part with the math overlaid", "build a lesson from this source document".