FrameVOX: Streamlining Agent-Driven Video Production via CLI

FrameVOX: A Video Production CLI for Agent-Made Social Videos

Manuel Bruña has released FrameVOX, a CLI designed for creating publish-ready videos using HTML compositions and TTS providers. The system integrates HyperFrames as its rendering engine to eliminate manual friction in the video production pipeline.

Why This Matters

Standard HTML-to-video workflows are fragile due to fragmented steps—spanning asset creation, voice conversion, and render linting—which frequently lead to failures when executed by AI agents. By replacing hidden setups and manual PCM-to-MP3 conversions with an explicit command path, FrameVOX transforms a high-friction manual process into a repeatable engineering workflow.

Key Insights

• Production Wrapper Architecture: FrameVOX acts as a thin layer over HyperFrames (2026), handling project scaffolding and TTS integration rather than replacing the renderer.
• Timing Synchronization: The system utilizes measured audio timelines from generated voice files rather than guessing text length to ensure precise video timing.
• Template Hierarchy: Implements a three-tier lookup order (Project -> User -> Builtin) allowing developers to scale from generic families like ‘promo’ or ‘studio’ to brand-specific global templates.
• Agent Integration: Includes a dedicated setup command (framevox setup) that installs skills for agent apps such as Claude Code, Cursor, and Codex.

Working Examples

Standard project initialization and render lifecycle.

npx framevox init my-promo --template minimal-mobile
npx framevox add-key gemini YOUR_GEMINI_KEY
npx framevox voice
npx framevox render

Voice script configuration supporting multi-scene delivery.

{
  "prompt": "Read with an energetic product launch tone:",
  "gap": 0.3,
  "scenes": [
    { "id": "hook", "text": "Your team schedule changed again." },
    { "id": "problem", "text": "Now three people are looking at three different plans." }
  ]
}

Practical Applications

• …Product Launch Reels: Using branded templates and Gemini TTS emotion tags (e.g., [excited]) to produce social demos; Pitfall: Guessing text length for timing instead of using generated audio timelines, leading to desynced visuals.
• …AI News Updates: Implementing automated scripts through agent skills in Cursor or Claude Code; Pitfall: Committing API keys to version control instead of using ~/.framevox/.env, risking security breaches.

References:

• https://github.com/tecnomanu/framevox