---
title: "Set up the speech recognition model provider in ScribeSTT"
description: "ScribeSTT needs one speech recognition model; confirm it is running in AI Model Center, get its instance name, then check the connection the system set up for you"
canonical_url: "https://docs.ardge.com/en/knowledge/scribestt/model-provider"
---
# Set up the speech recognition model provider in ScribeSTT

> ScribeSTT needs one speech recognition model; confirm it is running in AI Model Center, get its instance name, then check the connection the system set up for you

## Overview

ScribeSTT ships without a speech recognition model — transcription comes from one external **speech recognition model** (an ASR model), deployed in AI Model Center on the same ARX device.

That model corresponds to the **Model name** field on ScribeSTT's **Model connection** page. The **Base URL** and **API key** on that same page are filled in by the system at deployment time — neither is yours to enter.

That makes the setup short: confirm the model is running in AI Model Center, get its instance name, then go back to the **Model connection** page and check the **Model name** field.

## What you'll learn

- Confirm the speech recognition model's status in AI Model Center and get its instance name
- Check the connection the system set up on the **Model connection** page
- Confirm the setup with a real transcription

**Estimated time: about 10 minutes**

## Applies to

- ARX OS 1.0 and later
- ScribeSTT 1.3.0 and later

## Before you begin

- You are signed in to ScribeSTT as an administrator. **Model connection** sits in the **Administrator** group in the left navigation, and members don't see the page
- You have deployed a speech recognition model in AI Model Center on the same ARX device

<warning>

**The first time you open ScribeSTT, create the administrator account:**

Until the system is initialized, opening any page redirects to **Create administrator account**. Fill in **Username** (3–50 characters), **Email** (optional), **Password** (8–128 characters), and **Confirm password**, then click **Create and sign in** — you're signed in automatically.

</warning>

## What you'll need

Just one thing: the speech recognition model's **Instance Name**.

<table>
<thead>
  <tr>
    <th>
      Value
    </th>
    
    <th>
      What it's for
    </th>
    
    <th>
      Where to find it
    </th>
  </tr>
</thead>

<tbody>
  <tr>
    <td>
      <strong>
        Instance Name
      </strong>
    </td>
    
    <td>
      Checking the value in the <strong>
        Model name
      </strong>
      
       field on <strong>
        Model connection
      </strong>
    </td>
    
    <td>
      The model's <strong>
        Instance Name
      </strong>
      
       on the <strong>
        Model Instances
      </strong>
      
       page in AI Model Center
    </td>
  </tr>
</tbody>
</table>

**Base URL** and **API key** are filled in by the system for your deployment — there's nothing to collect for them.

## Step 1: Get the model's instance name from AI Model Center

1. Open **AI Model Center** on the same ARX device and go to the **Model Instances** page
2. Confirm the speech recognition model reads **Running**
3. Note its **Instance Name**

<warning>

**Transcription fails while the model isn't running:**

A model whose status isn't **Running** makes transcription fail even when **Model connection** is set up correctly, with "The speech recognition model isn't ready or isn't deployed. Confirm the model is running and try again." Start the model in AI Model Center first.

</warning>

## Step 2: Check the model connection page

1. Sign in to ScribeSTT and go to the **Model connection** page
2. Check that **Base URL** and **API key** both have values. **API key** is read-only and shows a masked version of the stored key
3. Check the **Model name** field against the **Instance Name** you noted in Step 1

If they match, the setup is done and you can go straight to Step 3.

If **Model name** points at a model you don't want to use, change it:

1. Open the **Model name** dropdown. The system probes the endpoint using **Base URL** and the stored key, and lists the models available there
2. Select the model you noted in Step 1
3. Click **Save**

The page shows "Model provider updated".

<note>

**The model list is probed live:**

**Model name** isn't a fixed list — its options come from probing **Base URL** the moment you open the dropdown. When nothing is found, it shows **No models found** — start by confirming the model reads **Running** in AI Model Center.

</note>

<tip>

**When you need to replace the key:**

**API key** is read-only and can't be edited directly. To swap in a different key, enter it in **Set new key** and save; left empty, the currently stored key stays in use.

</tip>

## Step 3: Confirm the setup with one transcription

1. Go to the **Transcriptions** page and click **New**
2. In the **Create new transcription** dialog, select a short audio file, select any **Output format**, and click **Create**
3. Watch that record's **Status** column in the list

The **Status** column moves through **Queued** and **Processing** to **Completed** — that means the model provider is set up correctly. For the full upload flow, see [Upload an audio or video file and get a transcript](https://docs.ardge.com/en/knowledge/scribestt/upload-and-transcribe).

## Troubleshooting

<table>
<thead>
  <tr>
    <th>
      Message
    </th>
    
    <th>
      Likely cause
    </th>
    
    <th>
      What to do
    </th>
  </tr>
</thead>

<tbody>
  <tr>
    <td>
      <code>
        The speech recognition model isn't ready or isn't deployed. Confirm the model is running and try again.
      </code>
    </td>
    
    <td>
      The model isn't running right now
    </td>
    
    <td>
      Confirm the model status is <strong>
        Running
      </strong>
      
       on the <strong>
        Model Instances
      </strong>
      
       page in AI Model Center
    </td>
  </tr>
  
  <tr>
    <td>
      <code>
        Can't reach the ASR model provider — check the base URL and the network.
      </code>
    </td>
    
    <td>
      ScribeSTT can't reach the model service
    </td>
    
    <td>
      Confirm the model service is running; if you have edited <strong>
        Base URL
      </strong>
      
       by hand, compare it against the endpoint AI Model Center shows
    </td>
  </tr>
  
  <tr>
    <td>
      <code>
        Unexpected response format from the ASR model provider.
      </code>
    </td>
    
    <td>
      The endpoint isn't OpenAI-compatible, or the <code>
        /v1
      </code>
      
       version prefix is missing
    </td>
    
    <td>
      ScribeSTT appends <code>
        /audio/transcriptions
      </code>
      
       to <strong>
        Base URL
      </strong>
      
      , so that URL has to end in <code>
        /v1
      </code>
      
      . If you have edited it by hand, put <code>
        /v1
      </code>
      
       back
    </td>
  </tr>
  
  <tr>
    <td>
      The dropdown shows <strong>
        No models found
      </strong>
    </td>
    
    <td>
      The probe found no models
    </td>
    
    <td>
      Confirm the model reads <strong>
        Running
      </strong>
      
      , and that <strong>
        Base URL
      </strong>
      
       points at AI Model Center's model endpoint
    </td>
  </tr>
</tbody>
</table>

## Next steps

With the model provider checked, start transcribing — see [Upload an audio or video file and get a transcript](https://docs.ardge.com/en/knowledge/scribestt/upload-and-transcribe).

## Related articles

- [What file formats and size limits does ScribeSTT support?](https://docs.ardge.com/en/knowledge/scribestt/formats-and-limits)
- [What to do when a transcription fails](https://docs.ardge.com/en/knowledge/scribestt/transcription-failures)
- [Call on-premises models on ARX through the API](https://docs.ardge.com/en/knowledge/arx-os/api-model-calls)


## Sitemap

See the full [sitemap](https://docs.ardge.com/sitemap.md) for all pages.