Gowajee API
  • Introduction
  • Getting Started
  • Speech to text
    • Models
    • Limitations
      • Rate Limit
      • Astro Model Limitation
      • Word Timestamp
      • Words Boosting
      • Audio Duration Limit
      • Handling Large Files
      • HTTP Multipart/Form-Data
      • Supported File Formats
    • Transcription
      • Synchronous API (Request-Response)
      • Asynchronous API (Webhook Notification)
        • Checking Job Status
    • Real-time Transcription
    • Speaker Separation
      • Multichannel
      • Diarization
    • Words Boosting
    • Raw Audio Format
  • Pricing
Powered by GitBook
On this page
  • Metrics Explained
  • Mapping Model with API
  • Example API Endpoints
  1. Speech to text

Models

PreviousGetting StartedNextLimitations

Last updated 1 month ago

Gowajee.ai offers three distinct models for speech-to-text conversion, each optimized for different use cases.

Below is an overview of our models.

Model
RTF
Description

Pulse

~0.17

Pulse strikes the best balance between accuracy and speed, making it ideal for applications where both performance and responsiveness are crucial.

Cosmos

~0.22

Cosmos is the most accurate model, and the speed is still acceptable.

✅ Recommended for most use cases.

Astro* (Deprecated)

~0.23**

Astro is a GPU-based model that is optimized for accuracy, providing a high level of precision for critical applications where accuracy is paramount. Update Oct 2024: Now, Cosmos has more accuracy than the Astro model.

Note:

(*) The Astro model has low availability as it requires GPU resources to compute. If you anticipate high demand and need to use the Astro model extensively, please to discuss your requirements and ensure adequate resource allocation.

(**) The Astro model is running on serverless GPU servers, The RTF of the Astro model does not include the cold-start time.

Metrics Explained

  • RTF (Real-Time Factor): This metric indicates the time taken by the model to process one second of audio (Lower is better). An RTF of 0.17 means the model takes 0.17 seconds to process one second of audio, highlighting the model's processing speed.

Mapping Model with API

This is how to map model names to their corresponding API paths. Each speech-to-text (STT) model in Gowajee has a specific API path for transcription requests. Use the correct path based on the model you intend to use.

Model
Value

Pulse

pulse

Cosmos

cosmos

Astro

astro

Example API Endpoints

Use the following API endpoints for transcription requests based on the chosen model:

  • Pulse Model:

    • Endpoint: /v1/speech-to-text/pulse/transcribe

  • Cosmos Model:

    • Endpoint: /v1/speech-to-text/cosmos/transcribe

  • Astro Model:

    • Endpoint: /v1/speech-to-text/astro/transcribe/async

contact us