# Connect AI voiceover providers

> Step-by-step - create an ElevenLabs, Sarvam or Deepgram API key, connect it to DemoVexa and choose the voice provider and default voice new videos use.

Source: https://demovexa.com/docs/getting-started/connect-voice

A **voice provider** records the AI voiceover of your videos: **ElevenLabs**, **Sarvam** or **Deepgram**. If you don't see one of them, it isn't switched on for your account yet.

A voice provider is optional. Without one, a video uses your own recording of the script or no voice (captions and music). See [Voice](/docs/videos/voice). The AI that writes and builds the videos is connected separately: see [Connect AI providers](/docs/getting-started/connect).

You connect a voice provider in the setup window the first time DemoVexa opens (the **AI voice** step), or later in **App settings → Voice providers**. The same guides are in the app: press **Guide** next to a provider.

## Which provider to pick

| Provider | Languages | Timing |
|---|---|---|
| **ElevenLabs** | 85 languages, with a large voice library and your own cloned voices | Words are highlighted exactly as they're spoken |
| **Sarvam** | Indian languages: Hindi, Bengali, Gujarati, Kannada, Malayalam, Marathi, Odia, Punjabi, Tamil, Telugu and Indian English (saved clones add Assamese) | Scene lengths come from the real audio; word highlighting is estimated |
| **Deepgram** (Aura) | English, Spanish, German, French, Dutch, Italian and Japanese | Scene lengths come from the real audio; word highlighting is estimated |

You can connect more than one. With an estimated provider, check the preview before you export.

## ElevenLabs

1. No ElevenLabs account yet? [Sign up for ElevenLabs](https://try.elevenlabs.io/4r5tshiikga0). Otherwise sign in at [elevenlabs.io](https://elevenlabs.io/app/sign-in). For videos you publish, choose a paid plan (**Starter** or higher): the free plan isn't for commercial use.
2. Click your profile at the bottom left → **Developers** → **API keys** ([direct link](https://elevenlabs.io/app/developers/api-keys)).
3. Click **Create API key** and name it **DemoVexa**. Leave it unrestricted, or if you restrict it, allow **Text to Speech**, **Voices** (read and write) and **User** (read, so DemoVexa can show the characters you have left). Click **Create**, then **Copy**: the key is shown only once.
4. In DemoVexa, paste it into **ElevenLabs API key**, click **Save**, then **Test connection**. It should say *Connected* with how many voices you have.

| What you see | What to do |
|---|---|
| *The key was rejected* | It was copied only partly, or deleted in ElevenLabs. Create a new key and paste it again. |
| Characters left doesn't show | The key can't read your account (the **User** permission). Voiceovers still work. |
| Out of characters | Upgrade in ElevenLabs, or wait for the monthly reset. The **Account** row in **App settings → Voice providers** shows what's left. |

## Sarvam

1. Sign in to the [Sarvam dashboard](https://dashboard.sarvam.ai/) and create an API subscription key with speech access.
2. In DemoVexa, paste it into **Sarvam API key**, click **Save**, then **Test connection**. The test makes a short Hindi sample, which uses a little of your Sarvam allowance.

The **Account** row links to your Sarvam dashboard, where you see your balance.

## Deepgram

1. In the [Deepgram console](https://console.deepgram.com/), create an API key with speech access.
2. In DemoVexa, paste it into **Deepgram API key**, click **Save**, then **Test connection**. The test makes a short English sample, which uses a little of your Deepgram allowance.

The **Account** row shows your Deepgram project balance when the key is allowed to read it.

## Choose your primary voice provider

Voice setup has three steps: **Connect Providers → Primary Voice Provider → Status**.

1. **Connect Providers.** Connect and test the providers you want, as above.
2. **Primary Voice Provider.** Pick the provider new videos use and its **Default voice**. Videos you've already made keep their voice.
3. **Status** shows each connection and the default voice, with a **Primary** badge.

A workspace can override this in **Workspace settings → Voice providers**, where **Language & narration** also sets the video language, the voice for each language and the voice for new videos. Each key is kept in your computer's keychain and only sent to its own provider.

## Next

- [Make your first video](/docs/getting-started/first-video)
- [Voice](/docs/videos/voice): pick a voice, record your own or go without
- [AI and voice integrations](/docs/settings/integrations)