rediumvex/ai-video-generator-claude

What it solves

This project provides a set of prompt engineering frameworks designed to turn Claude into a cinematic video director. It solves the problem of generic, low-quality AI video output by replacing vague prompts with production-grade, detailed instructions that Seedance 2.0 (on Higgsfield) can execute accurately.

How it works

Instead of simple templates, the project provides 10 specialized "skills" (system instructions) that encode cinematic vocabulary, the physics of attention (hooks), and precise timing. When a user provides a concept, these skills generate a 15-25 line prompt including:

• Timing and Hooks: Beat-by-beat breakdowns and scroll-stopping openers for the first 2 seconds.
• Cinematic Direction: Specific camera movements (e.g., dolly, orbit, whip pan) and lighting setups (e.g., three-point, neon contrast).
• Audio Integration: Timestamped sound design layers (ambient, impact, music, voice) synced to visual beats.
• Platform Optimization: Adjustments for aspect ratios and safe zones for TikTok, Reels, and Shorts.

Who it’s for

Content creators, founders, and marketers who use Claude and Seedance 2.0 to create high-conversion social media videos, SaaS product launches, and brand storytelling.

Highlights

• 10 Specialized Skills: Includes frameworks for Viral Hooks, SaaS Launches, Faceless Channels, and Luxury Aesthetics.
• Production-Grade Prompts: Generates detailed instructions including f-stops, rack focus, and specific camera phrasing.
• Detailed Timeline Architecture: Provides a second-by-second breakdown for clips between 4-15 seconds.
• Material Reference System: Guides the user on where to integrate external assets like images, videos, and audio files.