# OpenRouter APIs — Microsoft: MAI-Voice-2-Flash (microsoft/mai-voice-2-flash)

Guide for calling every confirmed OpenRouter API that serves `microsoft/mai-voice-2-flash`.

Model page: https://openrouter.ai/microsoft/mai-voice-2-flash
Create an API key: https://openrouter.ai/settings/keys

## Authentication

Send this header with every request:

- Authorization: Bearer $OPENROUTER_API_KEY

## Text-to-Speech API

Synthesize audio with `microsoft/mai-voice-2-flash` through OpenRouter's Text-to-Speech API.

Docs: https://openrouter.ai/docs/api/api-reference/tts/create-speech

### Endpoint

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

Headers:
- Content-Type: application/json

### Request fields (microsoft/mai-voice-2-flash)

- model: string (required) — `"microsoft/mai-voice-2-flash"`
- input: string (required) — text to synthesize
- voice: "en-US-Harper:MAI-Voice-2" | "es-MX-Valeria:MAI-Voice-2" | "fr-FR-Soleil:MAI-Voice-2" | "de-DE-Klaus:MAI-Voice-2" (optional) — provider-specific voice identifier
- response_format: "mp3" | "pcm" (optional) — output audio encoding; defaults to pcm

### Response

Success returns raw audio bytes, not JSON.

- `Content-Type` identifies the audio encoding, such as `audio/mpeg` or `audio/pcm`.
- `X-Generation-Id` identifies the billed generation.
- Write the response body directly to an audio file; do not call `response.json()`.

### Examples

#### Text to Speech

```bash
curl https://openrouter.ai/api/v1/audio/speech \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $OPENROUTER_API_KEY" \
  --output output.mp3 \
  -d '{
  "model": "microsoft/mai-voice-2-flash",
  "input": "The warm light flows across the OpenRouter office as the evening settles in and the routing never stops. One interface, hundreds of models. Ready when you are.",
  "voice": "en-US-Harper:MAI-Voice-2",
  "response_format": "mp3"
}'
```

#### Text to Speech 2

```bash
curl https://openrouter.ai/api/v1/audio/speech \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $OPENROUTER_API_KEY" \
  --output output.mp3 \
  -d '{
  "model": "microsoft/mai-voice-2-flash",
  "input": "Welcome to OpenRouter, one interface for hundreds of models. Curious how this voice sounds? This short sample was generated automatically to preview its quality.",
  "voice": "en-US-Harper:MAI-Voice-2",
  "response_format": "mp3"
}'
```

### Error differences

- 400 — malformed audio input or an audio option the routed provider does not support
- 502 — the upstream audio operation failed

## Errors

Failures return `{"error": {"code": <number>, "message": <string>}}` with the HTTP status:

- 400 — malformed request or an unsupported parameter
- 401 — missing or invalid API key
- 402 — insufficient credits
- 403 — spend limit reached, key disabled, or access blocked
- 404 — unknown model or no provider can serve the request
- 429 — rate limited; retry with backoff
- 502 — the operation failed upstream; failed generations are not billed

---

Canonical version of this document: https://openrouter.ai/microsoft/mai-voice-2-flash/llms.txt