Tara: Frontier Hindi + Hinglish Transcription

⚠️ Please use Chrome or Firefox — Safari has issues with audio recording/playback in Spaces.

Upload or record Hindi, Hinglish (Hindi–English code-switched), or English speech and get a transcript. Tara is a Whisper large-v3–based model with a dedicated mixed-code mode that writes English words in Latin script and Hindi in Devanagari — the way real Hinglish is written. Audio longer than 30s is transcribed in 30s windows.

Model card | Launch video | API access

Audio

Mode

Mixed-code is a safe default: pure-Hindi audio stays pure Devanagari, English words in mixed audio come out in Latin script. 'Hindi (pure Devanagari)' transcribes everything in Devanagari.

Hindi + mixed-code (recommended) Hindi (pure Devanagari) English Auto-detect language + mixed-code

Summary

Transcript

Examples