Bengaluru-based Murf AI has introduced Falcon 2, its latest text-to-speech model designed for real-time voice agents, as businesses increasingly explore conversational AI for customer service, sales and other voice-led interactions.
The new model is designed to address one of the key challenges in voice AI, latency. According to Murf AI, Falcon 2 can begin generating speech in under 100 milliseconds, helping reduce delays between a user's input and an AI agent's spoken response. The performance figures are company-reported.
Falcon 2 supports more than 150 voices across over 35 languages. Murf says the model can also handle code-mixed conversations, including interactions that switch between languages such as English and Hindi, an increasingly relevant capability for multilingual markets.
The model is aimed at enterprise applications including customer service, sales, call centres, recruitment and other conversational AI use cases. Murf's wider platform allows businesses to develop voice agents and integrate real-time speech capabilities into their existing workflows.
Murf said Falcon 2 can support up to 10,000 concurrent calls while maintaining stable latency. The company is positioning the technology for businesses that need to operate voice AI systems at larger volumes, where response speed and the ability to manage simultaneous conversations can affect the user experience.
The model has been developed specifically for real-time interactions rather than conventional text-to-speech applications such as pre-recorded narration. According to Murf, Falcon 2 was trained using conversational scenarios including customer calls, support interactions and booking conversations, with the aim of producing more natural pacing and emphasis.
The company also offers instant voice cloning, allowing businesses to create customised voices from short audio samples. Murf's enterprise offering includes data residency across multiple geographies and deployment options intended for organisations with security and infrastructure requirements.
Falcon 2 can be integrated with voice agent frameworks including LiveKit and Pipecat, giving developers the option to use Murf as the text-to-speech layer while combining it with separate speech recognition, large language model and telephony technologies.
The launch comes as voice is emerging as another interface for generative and agentic AI. Businesses are increasingly experimenting with AI agents capable of conducting spoken conversations and completing tasks, moving beyond text-based chatbots into applications such as customer support, appointment management and sales interactions.
For Murf AI, Falcon 2 expands its focus from AI-generated speech towards infrastructure designed for real-time enterprise conversations, with speed, multilingual capabilities and scalability positioned as the model's primary differentiators.
Disclaimer: This article may include information derived from interviews, press releases, public statements, research, company communications and other publicly available or third-party sources. Such material may be summarised, paraphrased or contextualised for journalistic and editorial purposes. All rights in third-party content remain with their respective owners.