Audio Speech

The /v1/audio/speech endpoint provides low-latency, streaming text-to-speech (TTS) synthesis across top-tier speech synthesis providers through a unified OpenAI-compatible API wire. By routing through nRouter, your infrastructure gains multi-provider failover, pre-call content moderation, transparent character-level billing at raw provider list prices, and comprehensive edge telemetry.

POST https://api.nrouter.ai/v1/audio/speech

Architectural Role & Lifecycle

When a client application submits text to /v1/audio/speech, nRouter executes a four-phase preflight lifecycle before emitting audio bytes:

1. Phase 1: In-Memory Authentication & ACLs: Verifies the virtual key hash (sk-nrouter-...), tenant organization status, and access permissions for the requested speech synthesis model.
2. Phase 2: Sliding-Window Rate Limiting: Enforces tenant-level and key-level requests-per-minute (RPM) and characters-per-minute ceilings without external cache bottlenecks.
3. Phase 3: Content Moderation & Prompt Injection Scoring: Evaluates the input text against safety policies. If flagged content or prompt injections are detected, the request is halted with HTTP 400 (x-nr-guardrails: blocked).
4. Phase 4: Character-Level Credit Reservation: Estimates the character volume and reserves credits against the tenant organization balance. If the preflight checks fail at Phase 3, zero credits are deducted or held.
5. Egress & Audio Streaming: Forwards the request to the upstream voice provider. As binary audio frames stream back, nRouter flushes the byte stream directly to the client while calculating exact cost settlement from actual characters rendered.

Request Parameters

The request body must be a JSON object containing the required synthesis parameters:

Request & Response Headers

Inbound Request Headers

Outbound Response Headers

SDK Code Examples

import { nRouter } from "@nrouter_ai/sdk";
import * as fs from "node:fs";

const client = new nRouter({
  apiKey: process.env.NROUTER_API_KEY,
});

const response = await client.audio.speech.create({
  model: "tts-1",
  voice: "alloy",
  input: "Welcome to nRouter, the multi-cloud enterprise AI routing gateway.",
  response_format: "mp3",
});

const buffer = Buffer.from(await response.arrayBuffer());
await fs.promises.writeFile("welcome.mp3", buffer);
console.log("Audio saved successfully to welcome.mp3");

Audio Codec & Format Specifications

nRouter supports six distinct audio encodings to accommodate varied client latency, fidelity, and bandwidth requirements:

Error Handling & Status Codes

Every error response emitted by nRouter returns a structured JSON payload accompanied by machine-readable error headers:

{
  "error": {
    "type": "gateway_error",
    "message": "Input text exceeds maximum allowed length of 4096 characters.",
    "code": "input_too_large"
  }
}

curl -X POST "https://example.com/v1/audio/speech" \  -H "Content-Type: application/json" \  -d '{    "model": "string",    "input": "string",    "voice": "string"  }'

"audio/mpeg bytes (rendered as an <audio> player)"