This repository provides a set of 10 specialized Claude "skills" designed to generate highly detailed, production-grade prompts for the Seedance 2.0 AI video generation platform on Higgsfield. It targets content creators, marketers, and businesses seeking to produce studio-quality AI videos efficiently, addressing the common issue of vague prompts yielding generic results. The benefit lies in transforming Claude into a cinematic director, automating the complex prompt engineering required for professional-looking video content, from viral hooks to SaaS demos and personal branding.

How It Works

The core approach involves users inputting a video concept, which the Claude skills then translate into a 15-25 line Seedance 2.0 prompt. This prompt meticulously details every second of the video, including precise timestamps, camera angles and movements (e.g., "dolly forward 3 feet, f/1.4"), specific lighting setups (e.g., "three-point, neon contrast"), and synchronized sound design layers. This method is advantageous because it encodes the "physics of attention" for viral hooks and utilizes a "cinematic vocabulary" that Seedance 2.0 understands, ensuring higher quality output than generic prompts.

Quick Start & Requirements

• Primary install: Clone the repository into your Claude skills folder: cd ~/.claude/skills && git clone https://github.com/rediumvex/ai-video-generator-claude.git ai-video-generator. An interactive installer (python install.py) is also available within the cloned directory.
• Prerequisites: Requires Claude Code and a pre-configured Claude skills directory (~/.claude/skills/ or ~/.claude/skills/gstack/).
• Links: GitHub repository (via clone URL), Roman Knox's Instagram (@theromanknox), Knox Community on Skool.

Highlighted Details

• Offers 10 distinct skills for generating specific video types: Viral Hook, Faceless Channel, Podcast Visual, SaaS Launch, Course Promo, Testimonial Story, Personal Brand, Luxury Aesthetic, Before-After, and AI Avatar.
• Each prompt includes engineered 2-second hooks, detailed timeline architecture, a camera movement encyclopedia, lighting/mood presets, a 4-layer sound design stack, and platform optimization.
• Claims to produce "studio-quality video" by focusing on attention physics, cinematic vocabulary, and sound-visual synchronization, directly from the developer's production pipeline.
• Each skill is described as a "deep engineering framework," not mere templates, providing comprehensive control over video elements.

Maintenance & Community

The project is built by Roman Knox, who focuses on AI products that generate revenue. Further community engagement and related tools can be found via his Instagram profile and the Knox Community on Skool.

Licensing & Compatibility

The project is released under the MIT license. Copyright is held by Roman Knox (2026). Standard MIT license terms apply, generally permitting commercial use and modification.

Limitations & Caveats

This tool is specifically designed for use with Seedance 2.0 on the Higgsfield platform, requiring users to have this environment set up. The effectiveness of the generated prompts is dependent on the capabilities of the underlying Seedance 2.0 engine and the user's conceptual input.