Tutorials: Vision

What You Will Build This tutorial demonstrates how to build an AI speech app with in-app voice cloning support. You can clone your favorite voice by supplying a reference audio of about 3 seconds. Here is a demo. Voice cloning example demonstrating reference and output voices Voice cloning example demonstrating reference and agent's output voices

Before You Start To begin, ensure that you meet these requirements and have the following credentials. Python 3.13 or a later version. An Apple Silicon Mac (recommended) or any modern laptop. Stream API credentials (for realtime audio and video communication). A HuggingFace Account and access token (HF_TOKEN). A Deepgram API key (for speech-to-text). A Google

When building voice AI applications, you have industry-leading cloud options for text-to-speech, such as Cartesia Sonic 3 and Grok TTS. For privacy and to avoid sharing your business’s data with these commercial text-to-speech (TTS) providers, your team may want to use free, open-source solutions that run locally on mobile and desktop devices. Continue reading to

Let’s build a restaurant reservation system to speak with a voice agent via a real-time phone call. The service will have three main features: Agent Outbound Call: The agent can act as both a customer helper and a restaurant assistant. For example, it can be configured as an AI restaurant employee that calls customers back

This step-by-step guide will help you build an AI front-desk receptionist that interacts with patients through conversations, assesses their conditions, and advises whether to visit a doctor or seek online medical advice. When an agent can see the patient’s condition in real time, it can make a smarter recommendation, saving patients an unnecessary trip to

Qwen 3.5 Small is a new family of lightweight, high-performance models from Alibaba (0.8B, 2B, 4B, and 9B parameters) that are now available on Ollama. These models support multimodal input, native tool calling, and strong reasoning, all while running efficiently on laptops, Macs, and even mobile/IoT devices. In this demo, the agent runs completely locally

Vision Agents has out-of-the-box support for the LLM services and providers developers need to build voice, vision, and video AI applications. The framework also makes it easy to integrate custom AI services — either by following a step-by-step guide or by vibe coding them using SoTA models. Let’s use Claude Opus 4.6 to create a

Most LLMs are great at thinking, but making them speak naturally is a different challenge. Gemini 3.1 Pro changes that. This new model from Google brings significantly improved reasoning, longer context, and better tool-use capabilities, making it one of the best choices (at the time of writing) for building conversational voice agents. In this guide,