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

# AssemblyAI

> Connect AssemblyAI to SentiVue Talk

AssemblyAI is a speech recognition platform, and in SentiVue Talk it provides the **transcription** of what the caller says.

## What you need

| Field   | Required |
| ------- | -------- |
| API key | Yes      |

## Getting your API key

1. Sign in to the [AssemblyAI dashboard](https://www.assemblyai.com/app).
2. Go to [API Keys](https://www.assemblyai.com/app/account) under your account.
3. Copy the key, or create a new one first.
4. Copy the key and store it somewhere safe before leaving the page.

Usage is billed against an account balance. If the balance goes negative, AssemblyAI rejects requests until you top up, so enable auto-pay or set a balance alert on the account you use for calls.

## Adding the connection

1. Go to **Connections** and choose **AssemblyAI**.
2. Paste the API key.
3. Save.

SentiVue Talk does not check the key with AssemblyAI when you save, so a wrong or expired key first shows up on a live call. Place a test call after saving.

## What you can select afterwards

Choose your model and language on [Model Configurations](/talk/configurations/inference-providers).

**Model** — SentiVue Talk offers AssemblyAI's real-time streaming models only, since a live call needs streaming transcription. See AssemblyAI's [models page](https://www.assemblyai.com/docs/speech-to-text/models).

**Language** — set it to match your callers. The English-only model accepts English variants; the other models cover English, Spanish, German, French, Portuguese and Italian, with regional variants to match your callers' accent. AssemblyAI's [streaming model comparison](https://www.assemblyai.com/docs/streaming/universal-3-pro/supported-languages) lists coverage per model.

## Completing the pipeline

AssemblyAI drives transcription only. You also need:

* A [language model](/talk/configurations/llm) to decide what the agent says
* A [voice provider](/talk/configurations/voice) to speak the replies

Each is a separate [connection](/talk/configurations/connections).

## Troubleshooting

<AccordionGroup>
  <Accordion title="The connection saved but calls produce no transcript">
    The key is not checked at save time, so a wrong or expired key only surfaces on a call. Copy the key again from **API Keys** in the AssemblyAI dashboard and confirm you are using the account you intend to bill.
  </Accordion>

  <Accordion title="Calls worked and then stopped">
    Check the account balance in the AssemblyAI dashboard. A negative balance blocks API requests until you top up. Enable auto-pay or set a balance alert to avoid this mid-month.
  </Accordion>

  <Accordion title="The language I want is not in the list">
    Streaming covers fewer languages than AssemblyAI's batch transcription, and the language options narrow further on the English-only model. Switch to a multilingual model to widen the list, or use a different [transcriber](/talk/configurations/transcriber) if your language is not covered.
  </Accordion>

  <Accordion title="Some calls lose transcription under load">
    AssemblyAI caps how many transcriptions run at once per account, and the cap on a free account is low. If you run many simultaneous calls, move to a paid account or ask AssemblyAI to raise the limit.
  </Accordion>
</AccordionGroup>