Tutorials: Vision (2)

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

The Qwen 3.5 Small model consists of 0.8B, 2B, 4B, and 9B parameters family of models on Ollama. They can be used for multimodal purposes, function/tool calling, and reasoning.

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

Gemini 3.1 Pro brings significantly improved reasoning, longer context, and better tool-use capabilities, making it one of the best choices (at the time of writing) for using it as an LLM in conversational voice agent workflows.

Cartesia Sonic 3 combines 200ms latency, emotions, multilingual support, and the ability to clone voices from input audio.

ElevenLabs delivers some of the most lifelike and expressive text-to-speech experiences in the commercial/consumer AI landscape. Characteristiscs of the speech synthesis model include natural intonation, emotion, and support for multiple languages.

Imagine pointing your desktop/mobile device's camera at everyday objects around to get live answers to any question. That's what the Kimi K2.5 vision model makes possible.

Scribe v2 Realtime is a low-latency speech-to-text model supporting 90+ languages with lower word error rate in accents.

Voice agents are getting better, but most text-to-speech pipelines still assume you're okay with cloud APIs, large models, and unpredictable latency. If you want fast, natural-sounding speech that runs entirely on your own hardware (no GPU, no network calls), you need a different approach. In this tutorial, you'll build a real-time AI voice agent that

The future of software is conversational and interactive. For developers, unlocking this frontier means moving beyond traditional text inputs to agents that can seamlessly see, hear, and speak. Our goal is to demonstrate a powerful, flexible architecture that achieves this. This allows us to build truly expressive, realtime-latency AI applications. To illustrate, consider our core

In this quick demo, we use Gemini 3 Flash to build a vision AI app under five minutes that watches your camera feed in real time and describes what it sees.

Integrate xAI's Grok-4 with Fish Audio's high-quality, expressive TTS and Deepgram's fast STT for a Siri-like app in Python.

Create an assistant that knows about GitHub branches, issues, pull requests, and more.

Drive-thru ordering is a deceptively hard real-time problem. Background noise, interruptions, fast-paced conversations, and the need for low-latency responses all push traditional voice systems to their limits. Modern speech-to-speech models change that equation by making natural, interruptible conversations possible without stitching together separate STT, LLM, and TTS pipelines. In this tutorial, you'll create a real-time

Google's Gemini 3, released November 18, 2025, gives you multimodal reasoning and tool-use for building AI apps. Let's combine it with Decart and other AI services to turn casual voice commands into artistic live video visual effects.

DeepSeek-V3.2 is accessible via OpenRouter. In less than 5 minutes, you can turn it into a vision-enabled math and physics assistant that not only solves problems but also explains its reasoning.

This tutorial demonstrates how to build an electronic device setup and repair assistant in Python with voice capabilities using Qwen3-VL hosted on Baseten. The assistant analyzes what a user shows on camera (like cables, ports, device components, or error states) and guides them step-by-step through setup or repair tasks. It's designed to reduce confusion during

Gemini models are supported in Vision Agents, the open-source Python framework for building voice and video AI applications. With Gemini 3, you will soon create an AI assistant that sees through your device's camera.

Large Language Models (LLMs) have been improving recently and are often used for building conversational applications for speech and transcription. From answering location-based questions to managing a work calendar, voice AI assistants are becoming an everyday part of both personal and professional life. In this tutorial, we'll take those same technologies a step further, using