Skip to main content
On the Configurations page, you can view, create, and manage all your transcription configuration templates. This guide walks you through creating a configuration using the user interface. A Configuration is a reusable template that defines how your audio/video files should be processed. For a detailed technical breakdown of the Configuration object and all its attributes, please see the Configuration Object Reference.
You can create up to 20 configurations. Each configuration is reusable across unlimited transcriptions.

​
Configuration List

On this page, you can see the list of your configurations. Each configuration shows its main features. You can filter and search for Configurations. You can also view the details of each one and delete the Configuration.

​
Creating a New Configuration

Click ”+ New Configuration” to open the configuration builder.

​
Configuration Builder Interface

The central panel contains all the settings to define your transcription and analysis workflow.

​
1. Configuration Details

These fields define the basic identification and functionality of your configuration.
FieldRequiredDescription
Identification TagYesA descriptive name for the configuration (max 255 characters).
SummaryNoEnable to have the AI automatically generate a summary (default: false).
Summary LanguageNoThe language for the summary (default: auto-detect). See Supported Languages.
Analytics LanguageNoLanguage for the Executive and Qualitative Analysis Report (default: auto-detect). See Supported Languages.
Use specific, descriptive tags like “Customer Support - Complaints” or “Sales Calls - Q4” instead of generic names like “Config 1”. This makes them easier to manage.

​
2. Custom Vocabulary

Improve transcription accuracy by adding an array of domain-specific words, phrases, or technical terms. Examples:
  • Product names: “Kubernetes”, “OAuth2”, “PostgreSQL”
  • Industry jargon: “blockchain”, “NFT”, “API Gateway”
  • Company terms: “Heify”, “TranscriptionBot”, “CloudSync”
  • Acronyms: “GDPR”, “SaaS”, “B2B”
A well-defined vocabulary significantly reduces transcription errors for terms that are key to your business.

​
3. Webhook Notifications

Configure URLs to receive real-time notifications about transcription status. Both fields are optional.
FieldDescription
Success URLReceives a POST request when transcription completes successfully.
Error URLReceives a POST request if transcription fails.
Example Success Payload: 
{
  "transcription_id": "f0e9d8c7-b6a5-4321-fedc-ba9876543210",
  "status": "COMPLETED",
  "configuration_id": "a1b2c3d4-e5f6-7890-1234-567890abcdef",
  "duration": 125.5,
  "completed_at": "2025-10-03T10:02:15.456Z"
}
Example Error Payload: 
{
  "transcription_id": "f0e9d8c7-b6a5-4321-fedc-ba9876543210",
  "status": "FAILED",
  "error": {
    "message": "Unsupported audio format",
    "code": 400
  }
}

​
4. Extraction Fields

Define structured data fields you want the AI to extract from your transcriptions. Learn more about Best Practices for Data Extraction for improve the quality of the extracted data. How to add fields:
  • Click “Create My First Field” or “Add Field”.
  • Fill in the field definition form:
    • Name: A unique identifier for the field (e.g., customer_id, sentiment).
    • Type: The data type: string, number, boolean, or array.
    • Description: Detailed instructions for the AI.
You can add a maximum of 20 extraction fields per configuration.
Alternatively, use the Quick Actions panel on the right to add pre-built extraction field templates with a single click.
After adding a quick action template, you can customize its name, type, and description to fit your exact needs.Learn more about extraction fields →

​
Best Practices for Data Extraction

To ensure accurate and consistent results, follow these guidelines when writing your field descriptions. 1. Use Bounded Responses Define a specific, limited set of possible values. This guides the AI and standardizes the output, making the data easier to process.
  • Good description: 
    "Classify the overall sentiment of the conversation. Must be one of: POSITIVE, NEGATIVE, or NEUTRAL."
  • Poor description: 
    "What is the sentiment?"
2. Provide Detailed Context and Examples The clearer your instructions, the better the results. Give the AI context and show it exactly what you’re looking for.
  • Good description: 
    "Extract the customer's order number. It's usually a 6-8 digit code 
starting with 'ORD-' or '#'. Examples: ORD-123456, #7894561."
  • Poor description: 
    "Get the order number."
Once created, a configuration cannot be edited. To make changes, you must create a new configuration with the desired settings and delete the old one if needed.

​
General Best Practices

Create configurations for common use cases and reuse them across many transcriptions and create a new one when the extraction needs are different.
Begin with 1-2 key extraction fields. Verify the quality of the results and refine their descriptions. Once you are satisfied, progressively add more fields to build complexity.

​
Troubleshooting

  • Issue: Fields consistently return null or incorrect values.
  • Common Causes & Fixes:
    • Vague Description: Make your instructions more explicit. Instead of “Get the name,” use “Extract the customer’s full name as mentioned in the conversation.”
    • No Examples: Add concrete examples of the format you expect in the field’s description.
    • Wrong Data Type: Ensure the selected data type (string, number, boolean, array) matches the information you are trying to extract.
  • Overly Complex Field: If you are trying to extract multiple pieces of information in one field, break it down into several simpler fields.

​
Need Help?

Support

Contact with us

Documentation

Explore the Sandbox guide