Phonic Speech AI Software

Phonic was started in 2024 by Moin Nadeem and Nikhil Murthy, two MIT graduates who had known each other for years. Nadeem was the eighth employee at MosaicML, which Databricks acquired in 2023. Murthy previously worked at NVIDIA, Intel, and Genesis Therapeutics, where he led a foundation model team.

The company's bet is that voice agents should not run on a transcript. A typical stack converts speech to text, passes it to a language model, and then synthesizes the answer. Phonic trains its own audio models end to end so the agent hears tone and hesitation directly, and it uses frontier language models as supervisor agents for reasoning and tool calls. Its team covers pre-training, inference work such as custom kernels, and forward-deployed engineers who work on site with customers.

In April 2025 Phonic raised a $4 million seed round led by Lux Capital. On October 29, 2026 it released Phonic-1 and announced a $27.5 million Series A led by First Round, with Lux Capital, Felicis, Definition, and Essence VC, for $31.5 million in total funding. The company says Phonic-1 ranked first on the Artificial Analysis speech-to-speech index and can act on a request before the caller finishes speaking.

Phonic says it handles calls on six continents. Ashby uses it for AI interviews in 25 countries, Maven AGI uses it in its Maven Voice product, and the company has published case studies with CloudTrucks and Rehearsals. It holds SOC 2 Type II certification and signs HIPAA BAAs.

1. Phonic-1 speech-to-speech model that takes audio in and returns audio out, without a transcript step

2. Tool calls that start while the caller is still speaking

3. Responses in about 600 milliseconds, by the company's measure

4. Trained on more than 50 languages, with mid-call language switching in the same voice

5. Frontier language models used as supervisor agents for reasoning and tool calling

6. No-code agent builder plus webhooks, WebSocket, and LiveKit integrations, with SDKs in TypeScript and Python

7. Phone numbers attached to agents, with selectable voices and background noise settings

8. Deployment on Phonic Cloud, single-tenant, VPC peering, or self-hosted for enterprise

9. SOC 2 Type II, HIPAA with BAA, a 99.99% uptime SLA, and SAML or OIDC single sign on

Phonic publishes per-minute pricing with no separate LLM pass-through or per-token charges. Starter costs $0.07 per minute with no minimum and includes 10 concurrent calls, 10 agents, and 3 workspace seats. Pro costs $0.065 per minute with a $200 monthly minimum, which covers about 3,100 minutes, and raises the limits to 50 concurrent calls, 50 agents, and 30 seats, with support over a private Slack Connect channel. Enterprise pricing is custom, with volume discounts, custom voices, on-premises or VPC deployment, BAAs for HIPAA, audit logs, and forward-deployed engineers. At launch Phonic offered $250 a month in free credits to anyone who signs up by November 30, 2026, and three months free for teams switching from another voice stack.

Ashby; Maven AGI; Rho; Assort Health; CloudTrucks; Rehearsals