Configure AI Call Transcription with OpenAI Service

Yeastar P-Series Cloud Edition supports AI call transcription powered by OpenAI Speech to Text service, transcribing the audio of a two-party call into readable text in real time. This topic describes how to configure AI call transcription with the third-party service on the PBX.

Tip: If you have already configured AI Voicemail Transcription powered by OpenAI Speech to Text service, you can enable Call Transcription and directly click Copy Settings from Voicemail Transcription to reuse the configuration, then modify the settings as needed.

Requirements

Yeastar P-Series Cloud Edition should meet the following requirements:

Item Requirement
Firmware 84.23.0.83 or later.
Subscription Subscribe to Enterprise Plan or Ultimate Plan to ensure the AI Transcription feature is available.

Prerequisites

Note: The prerequisites about GPT/Gemini LLM (Large Language Model) are required only if you want the PBX to automatically generate a call summary from the transcribed text after a call ends.
PBX network access
Make sure the Yeastar P-Series Cloud Edition can access the following domains to use corresponding services:
Note:

You can verify domain accessibility on the PBX (Path: Maintenance > Troubleshooting > IP Ping).

Service Domain
OpenAI Speech to Text service api.openai.com
GPT/Gemini LLM

According to your preferred LLM, the PBX must be able to access the following corresponding domains:

  • GPT LLM: api.openai.com
  • Gemini LLM: oauth2.googleapis.com and generativelanguage.googleapis.com
Third-party service account

You have prepared third-party service accounts and login credentials as follows:

Service Account
OpenAI Speech to Text service Prepare an OpenAI account with sufficient transcription minute quotas for OpenAI Speech to Text service, and obtain the username and password.
GPT/Gemini LLM According to your preferred LLM, the account must meet the following requirements:
  • GPT LLM: Make sure the OpenAI account have sufficient token quota.
  • Gemini LLM: Prepare a Google account with sufficient token quota, and obtain the username and password.

Procedure

Step 1. Create an API key for OpenAI Speech to Text service

To securely access the OpenAI Speech to Text service from the PBX, you must create an API key on OpenAI Platform first for the service API request authentication.

Note: The API key for the OpenAI Speech to Text service is identical to the one used to invoke GPT LLM (provided by OpenAI). If you want to use GPT LLM to automatically generate a call summary after a call ends, you can directly save the general API key created in this section for later use and skip Step 2.
  1. Log in to OpenAI Platform using your OpenAI username and password, go to API Keys.
  2. At the top-right corner of the API key list, click Create new secret key.
  3. In the pop-up window, create a new API key.

    1. In the Name field, enter a name to identify the API key.
    2. In the Project drop-down list, select the desired project.
    3. Keep the default All permission, and click Create secret key.

      A pop-up window appears, displaying the generated API key.

    4. In the pop-up window, click Copy to copy the API key and save it for later use.

Step 2. (Optional) Create an API key for Gemini LLM

Yeastar P-Series Cloud Edition allows you to invoke Gemini LLM (provided by Google) to automatically generate call summaries from transcribed text after calls end. To implement this feature, you need to create an API key using Google account to authenticate requests.

  1. Log in to Google AI Studio using your Google username and password, go to Dashboard > API keys.
  2. At the top-right corner of the API key list, click Create API key.
  3. In the pop-up window, create a new API key.

    1. In the Name your key field, enter a name to identify the API key.
    2. In the Choose an imported project drop-down list, select a desired project.
      Note: In the drop-down list, you can select an existing project, import a project, or create a new one as needed.
    3. Click Create key.

      A pop-up window appears, displaying the details of the generated API key.

    4. In the pop-up window, click Copy key to copy the API key and save it for later use.

Step 3. Configure AI call transcription on the Yeastar PBX

After you create authentication credentials for OpenAI Speech to Text service and LLM, you need to configure corresponding settings on the PBX to establish connections between the PBX and the two services.

  1. Log in to PBX web portal, go to AI > AI Toolbox.
  2. Turn on the switch of Call Transcription.
  3. Configure AI call transcription service.

    1. In the Service Type drop-down list, select Custom Service.
    2. In the Transcription Service Provider drop-down list, select OpenAI.
    3. In the API key field, paste the API key created on OpenAI platform.
    4. In the Transcription Model drop-down list, select your preferred transcription model supported by OpenAI Speech to Text service. For details on the differences between the transcription models, see the list of OpenAI Realtime & audio models.

    5. Configure the LLM service.

      Option Instruction
      Disable If you do not need the PBX to generate call summaries, select Disable in the LLM Provider drop-down list.
      OpenAI

      To use GPT LLM, complete the following settings:

      1. In the LLM Provider drop-down list, select OpenAI.
      2. In the API key field, paste the API key created on OpenAI Platform.
      3. In the GPT Model field, enter the model ID of your preferred GPT LLM.
        Note: You can access the list of GPT Models to check the model ID of your desired GPT LLM. For example, if you want to use GPT-5.4, enter the model ID gpt-5.4 in this field.
      Google

      To use Gemini LLM, complete the following settings:

      1. In the LLM Provider drop-down list, select Google.
      2. In the API key field, paste the API key created on Google AI Studio.
      3. In the Gemini Model field, enter the model code of your preferred Gemini LLM.
        Note: You can access the list of Gemini Models and go to introduction page of your desired Gemini LLM to check its model code. For example, if you want to use Gemini 3.1 Pro Preview, enter the model code gemini-3.1-pro-preview in this field.

  4. In the Language drop-down list, select the desired language to be detected and transcribed for call audio.

  5. In the Extension Scope for This Feature section, specify which extensions / extension groups / departments have access to the call transcription feature.
    • All Extensions: All extensions can use this feature.
    • Specific Extensions: Only selected extensions can use this feature.
  6. Click Save.

Result

  • The Transcription Connection Status displays Enable, indicating that the AI call transcription feature powered by OpenAI Speech to Text service is enabled. Call audio can be detected and transcribed into readable text in the specified language in real time via the OpenAI service.

    Note: You can configure the call transcription language and mode (either automatic or manual) for extensions individually as needed (Path: Extension and Trunk > Extension > AI > Call Transcription). For more information, see Configure AI Call Transcription for an Extension.

  • The LLM Connection Status displays Enable, indicating that the PBX is successfully connected to the configured LLM. The PBX will automatically generate call summaries from transcribed text after calls end via the LLM.