Models & Languages
Models Overview
Four models are available for speech-to-text transcription.
Available models
| Model | Description |
|---|---|
zero-indic | General-purpose model for Indian languages — Hindi, Tamil, Telugu, Kannada, Marathi, Bengali, and more. Use for most Indic + English audio. |
zero-med | Optimised for medical and clinical audio. Includes MedGemma-powered terminology correction for accurate medical transcription. |
zero-codeswitch | Specialised for code-switched speech (Hinglish, Tanglish, etc.). Uses Gemini-based restoration for accurate mixed-language transcription. |
zero-universal | 99-language Whisper-based model for broad multilingual coverage beyond Indic languages (English, European, Asian, and African languages). |