Build intelligent voice agents that handle customer support, drive sales, and deliver real customer outcomes. One API to tune and deploy natural-sounding agents that follow your procedures and use your tools.
A one-shot endpoint for spoken replies to uploaded audio, and a realtime WebSocket endpoint for live conversations. Both pay-per-use with no minimums.
Build real-time voice applications powered by Grok. Send recorded audio and get a voice and text response.
Build real-time voice applications powered by Grok. Stream audio and text bidirectionally via WebSocket for voice assistants, phone agents, and interactive voice systems.
Build full-duplex agents that listen while they speak, take interruptions in stride, and reach first audio in under a second. There's no STT → LLM → TTS chain to assemble, no vendors to keep in sync, and no latency lost between the hops.
Warm, expressive, and natural voice agents that can carry real conversations. Support for 25 languages with 25 stock voices and the ability to create new custom voices on the fly.
Hand agents your playbook and tools so it can reason through complex, multi-step requests just like a human would.
Send recorded audio over HTTP for a one-shot spoken reply, or open a WebSocket for a live conversation. The client libraries handle the queue protocol and the realtime framing.
import { fal } from "@fal-ai/client";
const result = await fal.subscribe("xai/grok-voice", {
input: {
audio_url: "https://example.com/question.wav",
prompt: "You are a concise, friendly voice assistant.",
voice: "aurora",
},
logs: true,
onQueueUpdate: (update) => {
if (update.status === "IN_PROGRESS") {
update.logs.map((log) => log.message).forEach(console.log);
}
},
});
// The agent's spoken reply (WAV) plus its transcript
console.log(result.data.audio.url);
console.log(result.data.text);Grok Voice is xAI's speech-to-speech model that supports real-time and one-shot voice agents. On fal.ai it comes as two endpoints: a one-shot API that takes an uploaded recording and returns the agent's spoken reply as a WAV file plus its transcript, and a realtime WebSocket API for live, bidirectional conversations.
Grok Voice 2.0 is widely regarded as the most intelligent voice model in the market. Public benchmarks from independent evaluators like Artificial Analysis show that Grok Voice leads frontier models on reasoning, agentic performance, speed, and overall ability to resolve customer issues.
$0.00083 per second of audio, about $0.05 per minute for both API endpoints. The model is pay-per-use with no minimums.
Install the fal.ai SDK (Python or JavaScript), grab an API key from your dashboard, and call the xai/grok-voice endpoint in a few lines of code, or open a WebSocket to xai/grok-voice/realtime with the realtime client. The API is serverless, so there is no infrastructure to set up. Check the API documentation for every available parameter.
Yes. Content generated through the fal.ai API can be used in commercial projects. Check fal.ai's terms of service for full details on usage rights and licensing.
Want to learn more about integrating Grok Voice into your workflow? Leave your details and our team will reach out.
