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

# Transcriber Setup

> Configure how speech is transcribed into text using STT providers like Deepgram or AssemblyAI.

The **Transcriber** tab determines how your assistant converts spoken audio into text using a Speech-to-Text (STT) engine.

To access this section:\
**Build → Assistant → Select Assistant → Transcriber**

***

## Why Transcription Matters

A fast, accurate transcriber ensures your assistant can understand callers clearly. The transcription engine directly affects recognition quality, especially across different languages, accents, and audio conditions.

***

## Supported Providers

CallHQ supports the following transcription providers:

* **Deepgram**
* *(More coming soon)*

Each provider offers its own balance of speed, accuracy, and pricing. You can select the one that best fits your use case and region.

***

## Configuration Options

Depending on the provider you choose, you may be able to configure:

### Language / Accent

Choose the primary language of your callers (e.g., English, Hindi, Hinglish). Some providers also support accent-specific models.

### Punctuation

Enable smart punctuation to improve readability (adds commas, periods, etc.).

### Profanity Filtering

Automatically mask or filter offensive language, if enabled by the provider.

### Confidence Threshold

Set a minimum confidence score for transcribed segments to be considered valid.

***

## Background Denoising

You can optionally enable **Background Denoising**, which filters out ambient noise (such as traffic, fans, or chatter) during transcription.

When toggled on:

* The transcriber applies noise suppression
* This results in cleaner, more accurate transcriptions, especially in noisy environments

***

## Best Practices

* Choose a provider optimized for your target language and audio environment.
* Enable punctuation for assistants that return transcribed messages via chat or CRM.
* Use confidence thresholds to reduce misinterpretation in noisy environments.

***