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

# Aero TTS

> Our text-to-speech model: natural voices in eleven Indian languages, streamed as they are made, in the console or through the API.

Aero TTS turns text into natural speech in **Bengali, English, Gujarati, Hindi, Kannada, Malayalam, Marathi,
Odia, Punjabi, Tamil and Telugu**. Mixed Hindi and English ("Hinglish") reads naturally too.

You can use it in two ways:

* **In the console:** open **Text To Speech** under **Model API** in the sidebar. Write a script, pick a voice, and generate. It plays as it is made.
* **Through the API:** stream speech into your own app, call center or voice agent. See [The API](https://docs.thinnest.ai/aero-tts/api.md) and
  [Integrations](https://docs.thinnest.ai/aero-tts/integrations.md).

## What is in the Text To Speech section

| Page | What it does |
| - | - |
| Text to Speech | One script, one voice, played back and downloadable |
| Conversation | Several speakers, each line in its own voice, joined into one clip |
| Voice Cloning | Make a voice of your own from a short recording |
| History | Every clip you made, to play or download again |
| Usage | Free characters left, and what Aero TTS cost over the last 30 days |
| Integrations | Code to use Aero TTS from LiveKit, Pipecat, the OpenAI SDK and more |
| Billing | Your wallet and plan, the same page as Settings → Billing |

## Voices

* **Catalogue:** about 40 voices, female and male, young to mature, from bright and upbeat to calm narrators. Every
  voice can speak every language. The picker shows the languages each one sounds most native in.
* **Listening first:** press play on any voice in the picker to hear a free sample.
* **Your own:** cloned voices appear under **My Cloned Voices**, and only in your workspace.
* **On your voice agents:** every Aero voice is on the **Premium** tab of an agent's voice picker. A call in one is
  billed at the Premium voice rate, like any other Premium voice.

## Language

* **Auto (the default):** reads the language from the script itself.
* **A specific language:** choose one in the Language picker, or pass a code such as `hi`, `ta` or `en` in the API.

## Billing

* **Free characters:** every workspace gets 1,00,000 characters once, about 100 minutes of speech, on any plan. They are used before your wallet, in the console and through the API, and whatever is left carries over when you upgrade. A developer's customers share the developer's characters.
* **How it is charged:** per character, from your wallet, on one statement line a minute for API use or one line a
  clip in the console.
* **The rate:** it depends on your plan, and is the same over HTTP and WebSocket. See the pricing page.
* **What counts as a character:** a letter, space or mark as written. A Devanagari letter is one character.
* **Failures are free:** a request that fails before any audio is sent costs nothing.
* **Stopping counts:** once audio has started, a clip counts all its characters, even if you stop it early.

> [!NOTE]
>
> **Aero TTS is billed in rupees,** so it is available to workspaces billed in INR. Workspaces of a white-label
> reseller's clients cannot use it yet.

## Limits by plan

| Plan | Requests speaking at once | Cloned voices |
| - | - | - |
| Pay as you go | 5 | 25 |
| Pro | 10 | 50 |
| Scale | 15 | 100 |
| Enterprise | Custom | Custom |

Notes on these limits:

* **One request over** your plan's limit gets a `429` with `Retry-After: 1`. Wait for one to finish and send it again.
* **Each request** may carry up to 4,000 characters.
* **A conversation** may run to 40 turns and 6,000 characters in all.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.