The speech-to-speech model company

Introducing Phonic-1, ranked #1 on the Artificial Analysis Speech to Speech Index.

Try Phonic-1 Book a demo

Or call +1 (781) 746 6426 to talk to a Phonic agent right now.

Trusted by teams shipping voice in production:

Ready

Click to start a live conversation

Why Phonic

Built for your most challenging conversations.

Hears the audio, not just the text.

Callers talk like they would to a person, because Phonic hears tone and hesitation, not just words.

Graph comparing Caller and Phonic waveforms over 600 ms, showing latency difference between the two.

Responds as fast as a person.

Phonic responds in 600 ms while other agents take a full second or more.

Orange silhouette of a person's head with wave lines and multilingual greetings in Arabic, Japanese, and Russian.

Fluent like a polyglot.

Serve callers in 50+ languages without breaking the voice or the conversation.

How it works

Natural speech without sacrificing on intelligence.

Highly natural speech. Sounds like a person, not a script. Phonic picks up the caller's tone, pace, and hesitation and responds in kind, in 600 ms.

Don't sacrifice intelligence. Phonic uses frontier models as supervisor agents to assist in tool calling and reasoning.

No more "let me check on that." Phonic-1 is the only speech-to-speech model that can call tools and take action while the caller is speaking.

Built for production

Drop into the stack you already run.

Phonic seamlessly plugs into webhooks, a held-open websocket, or LiveKit without rebuilds or refactors. Keep your current infrastructure as it is.

Start building Read docs

Phone number

TimezoneUsed to format system variables like {{system_time}}. To set the timezone per conversation, override system_conversation_timezone in template variables.

Voice

Background noise

Audio format

Audio speed1.2x

Welcome message

Hey there! How can I help you?

System prompt

You are a conversational Example variable here AI voice agent handling live customer support calls for a delivery company. Your goal is to resolve issues clearly while demonstrating empathy, emotional awareness, and calm confidence, especially when the customer is frustrated or stressed. Demonstrate empathy explicitly. Acknowledge the customer's feelings before moving to problem solving. Avoid sounding robotic or rushed. Reassure them that you are here to help.

import { Hono } from "hono";import { Webhook } from "svix"; const app = new Hono(); app.post("/webhooks/phonic", async (c) => { if (!process.env.PHONIC_WEBHOOK_SECRET) { return c.text("Bad Request", 400); } const wh = new Webhook(process.env.PHONIC_WEBHOOK_SECRET); const rawBody = await c.req.text(); try { const payload = wh.verify(rawBody, { "svix-id": c.req.header("svix-id") ?? "", "svix-timestamp": c.req.header("svix-timestamp") ?? "", "svix-signature": c.req.header("svix-signature") ?? "", }); console.log(JSON.stringify(payload, null, 2)); return c.text("OK", 200); } catch (error) { console.error("Failed to verify webhook:", error); return c.text("Bad Request", 400); }}); export default app;

// Also available in Python import { createNodeWebSocket } from "@hono/node-ws";import { Hono } from "hono";import type { WSContext } from "hono/ws";import { PhonicClient } from "phonic";import { phonicApiKey } from "./env-vars"; const app = new Hono();const phonicClient = new PhonicClient({ apiKey: phonicApiKey,}); const { injectWebSocket, upgradeWebSocket } = createNodeWebSocket({ app }); app.get( "/ws", upgradeWebSocket(() => { let phonicSocket: Awaited< ReturnType<typeof phonicClient.conversations.connect> > | null = null; let streamSid: string | null = null; const sendToTwilio = (ws: WSContext, data: unknown) => { ws.send(JSON.stringify(data)); }; return { async onOpen(_, ws) { phonicSocket = await phonicClient.conversations.connect(); phonicSocket.on("message", (message) => { if (streamSid && message.type === "audio_chunk") { sendToTwilio(ws, { event: "media", streamSid: streamSid, media: { payload: message.audio, }, }); } }); }, }; }),);

# Also available in Python import asyncioimport logging from dotenv import load_dotenv from livekit.agents import ( Agent, AgentServer, AgentSession, JobContext, cli, function_tool,)from livekit.plugins.phonic.realtime import RealtimeModel logger = logging.getLogger("phonic-agent") load_dotenv() class MyAgent(Agent): def __init__(self) -> None: super().__init__( instructions="You are a helpful voice AI assistant named Sabrina.", llm=RealtimeModel( voice="sabrina", audio_speed=1.2, ), ) @function_tool( description="Toggle a light on or off. Available lights are A05, A06, A07, and A08." ) async def toggle_light(self, light_id: str, state: str) -> str: """Called when the user asks to toggle a light on or off. Args: light_id: The ID of the light to toggle state: The requested state """ logger.info("Toggling %s to %s", light_id, state) return f"{light_id} is now {state}"

Build any workflow.

Use one SDK to run voice screening, dispatch coordination, patient intake, or anything else.

Audio-native by default.

Phonic owns the full audio path, routing speech into reasoning so you can build with fewer moving parts.

One model. One price per minute.

Replace vendor patchwork, multiple SLAs, and latency tradeoffs with one bill, one team, one stack.

Call a Phonic agent

Hear Phonic-1 on a real phone call.

Multilingual

Natively multilingual.

Phonic is trained on all the languages your customers speak. Since our agents sound like native speakers, callers speak as they normally would on a call.

Start building Read docs

Choose from 50+ languages

Play voice sample

Spanish - Sabrina

0:00 / 0:12

Enterprise ready

Production-grade voice. Secured for your team.

Trust center

Phonic Cloud

Default. Multi-tenant, multi-region. Elastic.

Single-tenant

Dedicated infrastructure in region of choice.

VPC peering

Connect Phonic to your private network.

Self-host (Enterprise)

Run Phonic in your own environment.

SOC 2 Type II certified

HIPAA + BAA available

99.99% Uptime SLA

SSO SAML / OIDC

News & Updates

Notes from the frontier of voice AI.

View all blog posts

Company

Introducing Phonic-1 and our $27.5M Series A

Today we're introducing Phonic-1, the most advanced speech-to-speech model for agentic tasks, and announcing our $27.5M Series A, led by First Round with participation from Lux Capital, Felicis, and Definition.

Phonic team

FAQs

Questions you ask before you sign up.

Chat with AI Assistant

What is Phonic?

Phonic is the speech-to-speech model company. We power calls across six continents for companies like Ashby, Maven AGI, Rho, and a Fortune 50 IT infrastructure company.

What types of conversations is Phonic best suited for?

Phonic is best suited for conversations where a voice agent needs to perform as well as a human agent. These are complex interactions that require strong reasoning, many tool calls, natural delivery, and sophisticated interruption handling.

What does Phonic charge?

Phonic charges one price per conversation minute, with no separate LLM pass-through or per-token charges. Starter is $0.07 / minute. Pro is $0.065 / minute with a $200 monthly minimum. Enterprise pricing is tailored to your needs.

How does Phonic compare with GPT Live?

Phonic is built for your most challenging conversations. Compared with GPT Live, Phonic delivers stronger intelligence, more natural voices, more sophisticated interruption handling, all at a lower latency.

How does Phonic compare with Vapi, Retell, Bland, and ElevenLabs?

These alternatives provide infrastructure that lets you assemble STT, LLM, and TTS models into cascaded voice agents. This was the first step in bringing voice AI into production, but these agents are slow, expensive, and brittle. We believe the end state of voice AI is speech-to-speech models, which is what we've been exclusively focused on since our founding.

Does Phonic offer white-glove implementation support?

Yes. Enterprise plans include access to forward-deployed engineers. Talk with our team about the implementation support your use case requires.