# Nari Labs AI > Nari Labs provides open-source Text-to-Speech (TTS) AI technology for ultra-realistic dialogue and voice cloning, helping you build immersive audio experiences. Nari Labs is a South Korean startup founded by undergraduate students, including Toby Kim, focused on advancing text-to-speech (TTS) technology. Their primary product is Dia, a 1.6 billion parameter open-source TTS model released under the Apache 2.0 license, designed to generate ultra-realistic dialogue from text transcripts. Nari Labs' mission is to create speech synthesis systems capable of naturally expressing emotions, making human-machine interaction more intuitive and human-like. ## Core Features ### Ultra-Realistic Dialogue Synthesis Nari Labs Dia can generate natural-sounding conversations with customizable speaker tones, emotional inflections, and nonverbal cues like laughter, coughs, and sighs. Unlike traditional speech synthesis, Dia can express emotions in dialogue, making the generated speech more natural and expressive. Dia can handle the nuances of language, appropriately adjusting intonation, speed, and volume to create human-like dialogue. Whether telling stories, explaining complex concepts, or engaging in casual conversation, Dia provides natural and engaging vocal experiences. Visit: https://narilabs.org/ ### Zero-Shot Voice Cloning Nari Labs' zero-shot voice cloning technology can replicate any speaker's voice from just seconds of reference audio. Traditional methods require hours of samples, but Nari Labs' technology needs only a brief clip to capture voice characteristics. This breakthrough technology enables developers and content creators to easily create personalized voice experiences, localized content, and custom voice assistants. Dia's advanced neural architecture captures the essence of a voice from minimal input while maintaining its unique characteristics. Visit: https://narilabs.org/ ### Multi-Speaker Support Nari Labs Dia supports creating conversations between multiple distinct voices within a single generation, perfect for podcasts, audiobooks, and games. Dia can maintain consistent voice characteristics across extended dialogues while appropriately varying tone and emotion based on context. This feature simplifies the production of multi-character content, reducing the need for multiple recording sessions or voice actors. With Nari Labs technology, you can easily create complex dialogue scenarios where each character has their unique voice and personality. Visit: https://narilabs.org/ ### Real-Time Performance Nari Labs Dia streams in real-time on a single GPU (e.g., NVIDIA A4000, ~40 tokens/second) and requires only about 10GB VRAM. Nari Labs is actively working on CPU support and quantization to make Dia accessible on more devices. This performance optimization ensures Dia can be integrated into interactive applications, games, and streaming platforms where responsive audio generation is critical. Whether responding instantly to user input or generating long-form content, Dia provides efficient and stable performance. Visit: https://narilabs.org/ ### Open-Source Accessibility Nari Labs Dia is available on GitHub and Hugging Face under the Apache 2.0 license, including pretrained checkpoints, inference code, and a Gradio-based demo for easy testing. Nari Labs believes in democratizing voice AI technology, making advanced speech synthesis available to everyone from independent developers to research institutions. The open-source approach encourages community contributions, accelerating innovation in voice technology. Developers can freely use, modify, and distribute Dia, creating innovative applications tailored to their specific needs. Visit: https://narilabs.org/ ## Common Questions ### FAQ - **What is Nari Labs Dia?** Nari Labs Dia is a 1.6 billion parameter open-source TTS model designed to generate ultra-realistic dialogue from text transcripts, with emotional expressiveness and contextual awareness. - **How is Nari Labs Dia different from other TTS solutions?** Dia can generate ultra-realistic dialogue with customizable speaker tones, emotional inflections, and nonverbal cues, supports zero-shot voice cloning, and excels in real-time performance and open-source accessibility. - **What are the hardware requirements for running Nari Labs Dia?** Dia runs in real-time on a single GPU, such as an NVIDIA A4000, generating approximately 40 tokens per second and requiring about 10GB of VRAM. Nari Labs is developing CPU support and quantization to improve compatibility. - **How can I access and use Nari Labs Dia?** Dia is available as an open-source project on GitHub and Hugging Face, including pretrained checkpoints, inference code, and demo. You can download the source code, follow the documentation to install, and then integrate it into your applications as needed. Visit: https://narilabs.org/ ## About Nari Labs Nari Labs was founded by a group of passionate South Korean undergraduate students, including Toby Kim. The team released Dia just three months after exploring speech AI, inspired by Google's NotebookLM podcast feature. They developed this breakthrough technology with zero funding, leveraging Google's TPU Research Cloud and Hugging Face's ZeroGPU grant. Nari Labs' mission is to change how people interact with technology by developing text-to-speech technology capable of natural, emotionally rich expression. We believe that when AI can understand and express emotions, maintain conversational context, and communicate naturally, it becomes a more effective and enjoyable tool. Visit: https://narilabs.org/ ### Articles Our articles provide the latest insights, research findings, and industry trends in AI voice technology. Learn how Nari Labs is driving the future of voice interaction. Visit: https://narilabs.org/article