PresentBuddy: a voice AI co-presenter for live talks, powered by Vocal Bridge

An AI partner that plans your presentation by conversation, generates the slides and animations, and speaks alongside you while you present, built on Vocal Bridge.

The problem with building and delivering a talk

Making a good presentation is two jobs, and both are slow. First you assemble the content, the slides, the diagrams, the animations that actually explain the idea. Then you have to stand up and deliver it, holding the thread, hitting your transitions, and not freezing when you lose your place. Most tools help with the first job and abandon you for the second. A slide editor does not jump in when you stall, and a script does not adjust to the pace of the room.

PresentBuddy treats both jobs as one continuous conversation. You plan the talk by speaking to the agent, it produces the assets, and then it stays with you on stage as a co-host, a backup, or a moderator, depending on how much help you want.

What PresentBuddy built

PresentBuddy is a voice AI co-presenter that works in two modes. In create mode you upload a document and plan the talk by voice, setting the format (slides or a continuous video), how much help you want on stage (CoHost, When Stuck, or Moderator), and the duration, and the agent generates the visuals and a structured script with handoff cues. In present mode it listens for those cues and participates at the level you chose, presenting whole sections, stepping in when you pause, or simply handling transitions.

The point is that a single spoken request turns into finished, animated content and a live presenting partner, with no slide editor in the middle. Voice fits because presenting is itself a spoken act, so a co-presenter you talk to beats one you type to.

How they use Vocal Bridge

PresentBuddy is built on Vocal Bridge's Voice for App surface, where the voice agent and the UI talk to each other on the same real-time data channel. When the agent decides to build an asset, it fires an action that the app renders into the Table Editor and the preview windows. When the presenter moves into Present mode, the agent receives the live transcript and watches for the handoff cues it wrote into the script.

The architectural split is clean. PresentBuddy owns the content engine, the document ingestion, the Manim animation and graph generation, the structured script with cues, and the support-level logic that decides when the agent speaks. Vocal Bridge owns the voice channel, the streaming text-to-speech, barge-in, and the sub-second transcript streaming that lets the agent track a live talk in real time.

import { useAgentActions } from "@vocalbridgeai/react";

const { onAction, sendAction } = useAgentActions();

// agent → app: a generated asset appears on the canvas
onAction("add_asset", ({ kind, title, payload }) => {
  setAssets((a) => [...a, { kind, title, payload }]);
});

// app → agent: presenter advanced a slide, tell the agent where we are
const handleAdvance = (slideId) => sendAction("slide_changed", { slideId });

Illustrative. The actual PresentBuddy implementation handles richer asset types and cue metadata.

What's hard about voice here

Live co-presenting is a timing problem, and the worst failure is talking over the person on stage.

• Tell a pause from a stall. The agent has to know a natural breath from a presenter who is genuinely stuck, and step in only when it should. Vocal Bridge's low-latency transcript streaming and barge-in give it the precise turn-taking that judgment needs.
• It has to keep up live. A co-host that lags is worse than none. Vocal Bridge's real-time speech is what lets it participate in the moment instead of trailing the talk.

That timing is the difference between a partner on stage and an interruption.

I would like to have you as a co-host for this.Mukul Pathak, who built PresentBuddy