Murf AI’s Falcon 2 voice model undercuts ElevenLabs at $0.01 a minute

Running real-time voice agents usually comes with a brutal trade-off: you either get robotic speech with awkward pauses, or you get crushed by API costs at scale.

Bengaluru-based startup Murf AI just launched Falcon 2 to tackle both problems at once. The new text-to-speech foundation model enters the market at $0.01 per generated minute, which is one-fifth of what top competitors like ElevenLabs charge for equivalent tiers.

Cutting costs by 80% while hitting 100ms latency

To put that pricing in perspective, generating one million minutes of audio on Falcon 2 costs $10,000, compared to roughly $50,000 on ElevenLabs Turbo or Flash.

Cost cuts usually raise quality concerns, but Falcon 2 actually scored higher than OpenAI's Realtime API on benchmarks hosted by independent tracker Artificial Analysis. The model responds in under 100 milliseconds, a critical threshold for making voice agents feel natural rather than lagging.

On the enterprise side, Murf AI claims the engine can handle up to 10,000 concurrent calls. It supports over 150 voices across 35 languages, allows data residency across 11 geographic regions, and supports deployments inside private corporate clouds. The company also noted that all voices are ethically sourced with explicit consent and ongoing royalties for voice actors.

Training on real customer chaos, not audiobooks

Most voice models stumble on the unglamorous details of customer support, like rattling off license plate numbers, street addresses, or multi-currency sums.

Murf AI COO Sneha Roy explained that the team trained Falcon 2 on real-world interactions, including actual support calls and booking workflows, instead of relying purely on clean, scripted audiobooks. As a result, the engine learns to understand sentence context and dynamically adjust cadence, emphasis, and pronunciation.

Founded in 2020 by IIT Kharagpur alumni and backed by Matrix Partners and Elevation Capital, Murf AI already serves over 10 million users and 300 enterprise clients, including Cisco, Pfizer, and Air France-KLM.

Why it matters

For businesses deploying voice bots in call centers, airlines, or fintech, Falcon 2 offers an immediate, massive cost reduction without obvious compromises on latency. Paying $0.01 per minute makes large-scale voice agent automation economically viable for teams that were previously priced out by premium incumbents.

Across the industry, this launch puts serious pricing pressure on dominant Western players like ElevenLabs and OpenAI. Voice AI is shifting from novelty demos to production unit economics, and Murf AI is proving that engineering teams outside Silicon Valley can beat industry darlings on both benchmark quality and infrastructure margins.

We think Murf AI's focus on practical friction points, like correctly reading account numbers and managing regional data residency, is exceptionally sharp. The main risk going forward is maintaining consistent quality and high uptime as 10,000-call concurrency gets battle-tested under messy real-world conditions.