receptron/mulmocast-cli

What it solves

MulmoCast addresses the limitation of traditional presentation tools (like PowerPoint) that were designed for humans, not AI. It provides an AI-native environment that allows generative AI to create rich, multi-modal content—such as videos, podcasts, slideshows, manga, and PDFs—from a single source of truth, keeping humans in the creative loop.

How it works

At the center of the system is MulmoScript, a JSON-based intermediate language that acts as a screenplay. LLMs generate this script to define the narrative structure, visuals, and audio. The MulmoCast CLI then processes this script through a pipeline:

1. Scripting: Generates a MulmoScript via interactive AI chat, web content, or local files.
2. Audio Generation: Uses TTS providers (OpenAI, ElevenLabs, Google) to create voiceovers.
3. Image Generation: Uses image models (OpenAI, Google, Replicate) to create visuals or renders slides via a Slide DSL (Tailwind CSS + Puppeteer).
4. Assembly: Combines these assets into final formats like MP4 videos with subtitles.

Who it’s for

It is designed for creators, researchers, and business professionals who want to transform reports, pitches, or stories into dynamic, multi-format content without manually building every asset.

Highlights

• Multi-Modal Output: Export content as video, podcast, slideshow, PDF, manga, or "Swipe Anime."
• Slide DSL: A structured JSON system for creating professional slides with 12 layouts and 10 content block types.
• AI-Native Workflow: Integrated tools for interactive script generation and automated translation.
• Usage Tracking: Built-in estimation and tracking of token, character, and cost consumption across multiple AI providers.
• Extensive Styling: Includes 100+ pre-designed CSS styles for markdown slides across various categories like business, tech, and nature.