Murf AI is stepping deeper into the increasingly competitive voice AI market with Falcon 2, a new text-to-speech model that aims to make real-time AI conversations faster, more natural and cheaper to operate.
The company is positioning Falcon 2 against established voice AI providers such as OpenAI and ElevenLabs. Its pitch centers on three areas that matter to businesses deploying voice agents at scale: latency, voice quality and cost.
Murf says Falcon 2 delivers 55 milliseconds of model latency and around 130 milliseconds of time-to-first-audio. The company also reports sub-100ms time-to-first-audio in its latest production and benchmark testing.
That speed matters because delays can make AI conversations feel unnatural. A voice agent that takes too long to respond can quickly become frustrating, particularly when it handles customer service, sales or other live interactions.
READ ALSO: Kano State Nigeria to Deploy 13,200 Computers in 44 ICT/CBT Centres
Terra Industries Hires Ex-SpaceX Director Ben MacWilliams as VP of Strategy
Kenyan Startups Raised $500M Before Collapse: Founders Now
Falcon 2 targets real-time voice applications rather than traditional text-to-speech alone. Murf’s API documentation highlights conversational AI, voice agents, virtual assistants and other applications where audio streams as the system generates it
Cost is another major part of Murf’s strategy.
The company prices Falcon 2 at 1 cent per minute, which it describes as an industry-leading price for real-time text-to-speech. For businesses processing large volumes of voice interactions, even small differences in speech-generation costs can become significant over time.
Murf also targets businesses that need voice AI across different markets. Falcon 2 supports more than 150 voices across 35 languages, with multilingual and code-mixed speech capabilities for conversations that move between languages.
The company is also making a strong argument around scale.
Murf says Falcon 2 can support up to 10,000 concurrent calls, while its enterprise offering includes data residency across 11 geographies and on-premise deployment for organizations that need greater control over their infrastructure and data.
Those capabilities put Falcon 2 squarely in the market for production voice agents.
A company building an AI receptionist, customer service agent or sales assistant needs more than a voice that sounds human. It also needs the system to respond quickly, handle large call volumes and keep operating costs under control.
Murf is using benchmark results to strengthen its case against competing models.
In its published latency testing, Murf compared Falcon 2 with real-time models from ElevenLabs, OpenAI, Cartesia and Deepgram. The company says Falcon 2 recorded the fastest time-to-first-audio across most of the regions it tested, using a third-party geo-distributed API relay across 33 locations.
Murf’s current site also points to Artificial Analysis, an independent benchmarking platform, where it says Falcon 2 ranks above real-time voices from ElevenLabs, OpenAI and xAI for naturalness. That provides an independent reference point, although readers should consider the benchmark within Artificial Analysis’ own methodology rather than treat it as a universal measure of voice quality.
The competitive landscape is becoming increasingly crowded as voice AI moves beyond demonstrations and into everyday business operations.
OpenAI has invested heavily in real-time voice capabilities, while ElevenLabs has built a strong position around AI-generated speech and voice agents. Other companies, including Cartesia and Deepgram, also compete for developers building real-time conversational systems.
That makes Murf’s approach particularly interesting.
Rather than competing only on how realistic an AI voice sounds, the company targets the infrastructure behind those conversations. Latency, pricing, multilingual support, scalability and deployment options could become just as important as voice quality as companies move from experimenting with voice agents to deploying them at scale.
Falcon 2 also reflects a broader change in the voice AI market.
Businesses increasingly want AI agents that can answer calls, qualify leads, schedule appointments, provide customer support and handle routine conversations without human intervention. As those systems become more common, the cost and speed of every interaction will matter more.
A model that responds almost instantly while keeping generation costs low could therefore become valuable even if it does not lead every measure of voice quality.
That is the market Murf is trying to capture with Falcon 2.
The company is betting that the next phase of voice AI will depend less on impressive demonstrations and more on production performance. If Falcon 2 can deliver its claimed combination of low latency, natural speech and low operating costs at scale, Murf will have a stronger case for competing with the industry’s established voice AI leaders.
The voice AI race is no longer simply about making machines sound human. It is becoming a race to build voice systems that can respond instantly, understand real-world conversations and operate cheaply enough to handle millions of interactions.
