kunchenguid/autopreso

Realtime speech to presentation. Let the whiteboard whiteboard itself.

What it solves

autopreso allows presenters to create and update a visual whiteboard in real time using only their voice. It eliminates the need to manually click, draw, or rearrange elements on a slide deck or canvas during a presentation, enabling a hands-free experience.

How it works

The system operates as a local web app that integrates a live Excalidraw canvas with a listening AI agent. The process follows this pipeline:

  1. Speech-to-Text (STT): Audio from the browser microphone is streamed to a transcription model (either local Moonshine or OpenAI Realtime).
  2. Agent Processing: Transcripts are sent to an LLM agent (via OpenAI, Codex, or Ollama) that interprets the speech.
  3. Live Updates: The agent uses tool calls to draw, label, and rearrange elements on the Excalidraw scene in real time.

Users can start in a "staging" mode to set up seed content and instructions, then switch to "live" mode to let the agent drive the canvas based on their speech.

Who it’s for

Presenters, educators, and speakers who want dynamic, visually-driven presentations without the need for a manual slide deck or a clicker.

Highlights

  • Hands-free operation: Speech drives the agent that edits the whiteboard.
  • Flexible Model Support: Compatible with OpenAI, Codex, and Ollama.
  • Local-first option: Can run entirely locally using Moonshine for transcription and Ollama for the agent.
  • Integrated Canvas: Built on top of Excalidraw for high-quality whiteboard rendering.
  • Hybrid Workflow: Includes a staging mode for seed content and a live mode for real-time updates.

Related

  • Project
    yctimlin/mcp_excalidraw

    A local Excalidraw canvas that AI agents can control via CLI, MCP, or REST, enabling programmatic creation, inspection, and version‑controlled export of diagrams.

  • Dispatch
    VoiceDraw: AI-Powered Voice-to-Diagram System Design Tool

    VoiceDraw is a tool that automatically generates system design diagrams and architectural notes from spoken conversation, enabling developers to visualize system architecture in real-time during discussions or interviews.

  • Project
    netease-youdao/LobsterAI

    A desktop AI agent that automates office workflows by interacting with local files, terminal commands, and browser workflows through a dedicated runtime engine.

  • Project
    sligter/LandPPT

    An AI-driven presentation platform that transforms topics or documents into professional, editable PPTs with automated outlining, research, imaging, and video export.

  • Project
    icip-cas/PPTAgent

    An AI agent that automates the creation of professional presentations, transforming prompts and documents into fully formatted PPTX files using a dedicated toolset and fine-tuned models.