Skip to main content

When to Use Custom Models

Custom models in Scriptix allow you to train specialized speech recognition models for improved accuracy in specific domains.

What are Custom Models?​

Custom models are specialized transcription models trained on your specific:

  • Domain-specific audio data
  • Industry terminology
  • Speaker characteristics
  • Audio environments

When to Consider Custom Models​

High-Volume Specialized Content​

Custom models may be beneficial when you have:

  • Large volumes of domain-specific audio
  • Consistent specialized terminology
  • Recurring speakers or audio environments
  • Need for high accuracy in specific domains

Industry-Specific Use Cases​

Medical & Healthcare:

  • Medical consultations and dictations
  • Clinical documentation
  • Medical education content
  • Specialized medical terminology

Legal:

  • Depositions and legal proceedings
  • Court recordings
  • Legal consultations
  • Legal terminology

Technical & Business:

  • Technical meetings and presentations
  • Industry-specific content
  • Corporate communications
  • Specialized business terminology

Decision Factors​

Consider Custom Models If:​

  • High Volume: Regular, ongoing transcription needs
  • Specialized Vocabulary: Domain-specific terms not recognized well by base models
  • Consistent Content: Similar audio characteristics across recordings
  • Accuracy Requirements: Need for high accuracy in specialized domains
  • Long-term Use: Ongoing use case justifies training investment

Use Base Models + Glossaries If:​

  • Limited Specialized Terms: Only specific names or terminology
  • Lower Volume: Occasional transcription needs
  • General Content: Standard speech patterns
  • Quick Implementation: Need immediate solution
  • Budget Constraints: Limited resources for custom training

Alternatives to Custom Models​

1. Glossaries​

Glossaries are simpler to implement and effective for:

  • Product and company names
  • Limited specialized terminology
  • Quick accuracy improvements
  • Lower cost solution

2. Base Model Selection​

Try different base models:

  • General model for standard speech
  • Medical model for healthcare content
  • Legal model for legal proceedings
  • Technical model for technical content

3. Audio Quality Improvement​

Better recording quality improves accuracy:

  • Use quality microphones
  • Reduce background noise
  • Optimize recording environment
  • Consistent audio setup

Getting Started​

Progressive Approach​

  1. Start with base models - Evaluate accuracy with standard models
  2. Add glossaries - Include specialized terminology
  3. Assess results - Measure accuracy improvements
  4. Consider custom models - If additional accuracy is needed

Prerequisites​

Before training custom models:

  • Sufficient training data available
  • Consistent audio quality
  • Clear business case
  • Resources for training and maintenance

Accessing Custom Models​

Custom model features depend on your organization's plan and configuration.

Contact your organization administrator for information about:

  • Custom model availability
  • Training requirements
  • Implementation process
  • Plan requirements and pricing

Next Steps​

Note: Custom model features and availability vary based on your subscription plan. Contact your organization administrator for details about custom model access and capabilities.