<!DOCTYPE html>

<html lang="en">

<head>

<!-- By Jim Salsman, April 2025. Released under the free MIT License. -->

<meta charset="UTF-8">

<title>Gemini Live Voice to Text Realtime: Enter API Key</title>

<meta name="description" content="Gemini Live voice-to-text realtime stream">

<meta name="viewport" content="width=device-width, initial-scale=1.0">

<meta property="og:title" content="Gemini Live Voice to Text Realtime">

<meta property="og:description" content="Gemini Live voice-to-text realtime stream">

<meta property="og:image" content="https://live.talknicer.com/screenshot.png">

<meta property="og:url" content="https://live.talknicer.com/">

<meta property="og:type" content="website">

<meta property="og:site_name" content="live.talknicer.com">

<meta name="twitter:card" content="summary">

<meta name="twitter:title" content="Gemini Live Voice to Text Realtime">

<meta name="twitter:description" content="Gemini Live voice-to-text realtime stream">

<meta name="twitter:image" content="https://live.talknicer.com/screenshot.png">

<style>

body {

background-color: #afa;

font-family: sans-serif;

text-align: center;

}

a { text-decoration: none; }

.description {

width: 80%;

margin: 0 auto;

text-align: left;

}

</style>

</head>

<body>

<h2>Gemini Live Voice to Text Realtime</h2>

<p>Enter your (free!) Gemini API key from:

<a href="https://aistudio.google.com/apikey" target="_blank"

style="text-decoration: none;">aistudio.google.com/apikey</a></p>

<form method="POST">

<input type="password" size="33" name="api_key" required>

<button type="submit">Run</button>

</form>

<br>

<p class="description">

This is a web application that allows you to interact with

Google's Gemini 3.8 Live large language model using your

voice in real-time. It's based on Google's <tt>js-genai</tt>

API to run entirely in the browser, providing a seamless,

voice-driven experience. Key features include Google Search,

calculations and matplotlib plots using Gemini Flash's managed

Python execution, silent streamed response transcription,

image upload (including from the camera

on mobile), typed text, markdown and LaTeX output, and

conversation preservation across Stop/Start Listening.

The <a href="https://github.com/jsalsman/gemini-live"

target="_blank">source code is free on GitHub</a> under the

MIT License.

<em>Voice to text interactions are advantageous because most

people read about twice as fast as synthetic speech typically

talks, while text can be skimmed. Interrupting speech leaves

what might have been said as a frustrating mystery. Furthermore,

you can copy from text but not speech.</em>

</p>

<br>

<p class="description">

<b>Privacy policy:</b> Your API key is saved in a cookie at this domain.

Flask receives the initial form and subsequent cookies but does not

persist the key in server storage. Browser JavaScript sends your key

directly to Google's Gemini Live and Flash APIs. Execution tasks and

their data are sent to Google; generated Python runs in Google's managed

environment. Connection or model errors keep your key. Gemini usage is

subject to your project's quota and billing. You can also fork the

code on GitHub and run it on your own server.

</p>

<br>

<img src="/screenshot.png" width="60%" alt="screenshot">

<p>

By <a href="mailto:jim@talknicer.com"

>Jim Salsman</a>, April 11, 2025.

<a href="https://paypal.me/jsalsman">Donate.</a>

</p>

</body>

</html>