Telugu is spoken by roughly 80 million people, but remains severely under-resourced in the ASR landscape. whisper-large-v3-te is our fine-tuned version of OpenAI’s Whisper Large v3, trained specifically for Telugu speech recognition.

Why Telugu?

Standard Whisper Large v3 performs reasonably on high-resource languages but degrades significantly on Telugu. The fine-tune targets conversational and broadcast Telugu, covering a range of dialects and recording conditions.

Evaluation

We evaluate using both standard WER and SemanticWER — our own metric that accounts for semantically equivalent transcriptions that differ in surface form. This matters for agglutinative languages like Telugu where the same utterance can have multiple valid transcriptions.

The model is available on HuggingFace under Apache 2.0. Inference examples and evaluation code are on GitHub.