> ## Documentation Index
> Fetch the complete documentation index at: https://developers.telnyx.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Alternative Providers

> Use third-party text-to-speech providers — AWS Polly, Azure, ElevenLabs, MiniMax, Rime, Resemble, Inworld, Fish Audio — through a single unified Telnyx Voice API.

Telnyx supports routing TTS requests to alternative providers. Each has its own voice format, audio behavior, and voice settings.

<Info>
  All provider-specific settings below are sent via the `voice_settings` object in the [initialization frame](/docs/voice/tts/websocket-streaming#2-streaming), **not** as query parameters. Only the [connection parameters](/docs/voice/tts/websocket-streaming/parameters) (`voice`, `provider`, `model_id`, `voice_id`, `language`, `socket_id`, `disable_cache`) are set on the URL.
</Info>

***

## AWS Polly

**Voice format:** `aws.Polly.<Engine>.<VoiceId>` — e.g., `aws.Polly.Generative.Lucia`

**Audio:** Concatenated (full audio in a single chunk). Format controlled via `output_format` and `sample_rate`.

**Engines:** `standard`, `neural`, `generative`, `long-form`. Engine can also be parsed from a hyphenated voice ID suffix (e.g., `Lucia-longform`).

### Voice Settings

| Field           | Type   | Default      | Description                                     |
| --------------- | ------ | ------------ | ----------------------------------------------- |
| `language_code` | string | —            | BCP-47 language code                            |
| `engine`        | string | `"standard"` | `standard`, `neural`, `generative`, `long-form` |
| `text_type`     | string | `"text"`     | `text` or `ssml`                                |
| `output_format` | string | —            | Audio format override                           |
| `sample_rate`   | string | —            | Sample rate                                     |
| `lexicon_names` | array  | —            | Pronunciation lexicon names                     |

***

## Azure Speech

**Voice format:** `azure.<VoiceId>` — e.g., `azure.en-US-AvaMultilingualNeural`

**Audio:** Concatenated. Default: `audio-24khz-160kbitrate-mono-mp3`. Configurable via `output_format`.

No model ID segment — Azure voices are flat identifiers. Default voice: `en-US-AvaMultilingualNeural`.

### Voice Settings

| Field           | Type   | Default                              | Description                  |
| --------------- | ------ | ------------------------------------ | ---------------------------- |
| `language_code` | string | `"en-US"`                            | Language code                |
| `output_format` | string | `"audio-24khz-160kbitrate-mono-mp3"` | Azure output format string   |
| `text_type`     | string | `"text"`                             | `text` or `ssml`             |
| `effect`        | string | —                                    | `eq_car` or `eq_telecomhp8k` |
| `gender`        | string | —                                    | `Male` or `Female`           |

***

## ElevenLabs

**Voice format:** `elevenlabs.<Model>.<VoiceId>` — e.g., `elevenlabs.v3.Adam`

**Audio:** Direct relay — frames are forwarded to ElevenLabs' upstream WebSocket. Text is not buffered at sentence boundaries. Audio format is determined by ElevenLabs and is not configurable through Telnyx.

### Voice Settings

Settings are relayed directly to the upstream API:

| Field               | Type    | Description                  |
| ------------------- | ------- | ---------------------------- |
| `model_id`          | string  | ElevenLabs model ID          |
| `language_code`     | string  | Language code                |
| `stability`         | float   | Voice consistency (0.0–1.0)  |
| `similarity_boost`  | float   | Voice clarity (0.0–1.0)      |
| `style`             | float   | Style exaggeration (0.0–1.0) |
| `use_speaker_boost` | boolean | Speaker boost toggle         |

***

## Minimax

**Voice format:** `minimax.<Model>.<VoiceId>`

**Audio:** Streamed. Default: MP3 (32kHz, 128kbps). Set `response_format` to `"pcm"` for raw PCM (16kHz, mono).

Supports voice cloning (scoped to your organization).

### Voice Settings

| Field             | Type    | Default | Description       |
| ----------------- | ------- | ------- | ----------------- |
| `speed`           | float   | —       | Playback speed    |
| `vol`             | float   | —       | Volume            |
| `pitch`           | integer | —       | Pitch adjustment  |
| `language_boost`  | string  | —       | Language emphasis |
| `response_format` | string  | `"mp3"` | `mp3` or `pcm`    |

***

## MurfAI

**Voice format:** `murfai.FALCON.<VoiceId>`

**Audio:** Default: MP3, 16kHz, mono. Configurable via `format` and `sample_rate`.

### Voice Settings

| Field         | Type    | Default | Description                         |
| ------------- | ------- | ------- | ----------------------------------- |
| `pitch`       | integer | 0       | -50 to 50                           |
| `rate`        | integer | 0       | -50 to 50                           |
| `format`      | string  | `"MP3"` | Output format                       |
| `sample_rate` | integer | 16000   | 8000, 16000, 24000, 44100, or 48000 |

***

## Rime

**Voice format:** `rime.Coda.<VoiceId>` or `rime.ArcanaV3.<VoiceId>`

**Models:** Coda (flagship, recommended) and ArcanaV3 (previous flagship). Coda offers sub-100ms latency, 184 voices, and supports English, Spanish, French, Portuguese, German, and Japanese.

**Audio:** Default: MP3, 24kHz.

### Voice Settings

| Field             | Type    | Default | Description       |
| ----------------- | ------- | ------- | ----------------- |
| `voice_speed`     | float   | 1.0     | Speech rate       |
| `response_format` | string  | `"mp3"` | Audio format      |
| `sampling_rate`   | integer | 24000   | Sample rate in Hz |

***

## Resemble

**Voice format:** `resemble.Turbo.<VoiceId>`

**Audio:** Default: MP3, 48kHz. WAV defaults to 16kHz.

### Voice Settings

| Field         | Type   | Default                          | Description                                |
| ------------- | ------ | -------------------------------- | ------------------------------------------ |
| `precision`   | string | `"PCM_32"`                       | `PCM_16`, `PCM_24`, `PCM_32`, `MULAW`      |
| `sample_rate` | string | `"48000"` (mp3), `"16000"` (wav) | 8000, 16000, 22050, 32000, 44100, or 48000 |
| `format`      | string | `"mp3"`                          | `wav` or `mp3`                             |

***

## Inworld

**Voice format:** `inworld.<Model>.<VoiceId>`

**Models:** `inworld-tts-1.5-mini` (alias `Mini`, faster), `inworld-tts-1.5-max` (alias `Max`, higher quality), and `inworld-tts-2` (alias `TTS2`, latest generation; supports the `delivery_mode` setting). Defaults to `inworld-tts-1.5-mini` if omitted.

**Audio:** Default: MP3, 24kHz.

### Voice Settings

| Field           | Type    | Default | Description                                                             |
| --------------- | ------- | ------- | ----------------------------------------------------------------------- |
| `encoding`      | string  | `"MP3"` | `MP3` or `LINEAR16`                                                     |
| `sample_rate`   | integer | 24000   | Sample rate in Hz                                                       |
| `language_code` | string  | —       | Language code                                                           |
| `delivery_mode` | string  | —       | `STABLE`, `BALANCED`, or `CREATIVE`. Only supported by `inworld-tts-2`. |

***

## Fish Audio

**Voice format:** `FishAudio.<Model>.<VoiceId>`

**Audio:** Streamed. Default: raw PCM at 24 kHz. Configurable via `format` and `sample_rate` in voice settings.

Telnyx exposes a **hand-vetted shortlist** of Fish Audio voices. Only curated catalog voices are accepted for synthesis. Fish Audio is **cross-lingual** — any voice speaks any language from the input text.

### Voice Settings

| Field         | Type    | Default | Description                                                                                                 |
| ------------- | ------- | ------- | ----------------------------------------------------------------------------------------------------------- |
| `format`      | string  | `"pcm"` | `mp3`, `wav`, `pcm`, `opus`                                                                                 |
| `sample_rate` | integer | `24000` | Valid values depend on format: pcm/wav — 8000, 16000, 24000, 32000, 44100; mp3 — 32000, 44100; opus — 48000 |

No speed or pitch controls. Emotion and tone are controlled inline in the text via bracket markers on S2 models — see the [Fish Audio provider page](/docs/voice/tts/providers/fishaudio#emotion-tags) for syntax and examples.