Model

Whisper Medium TR

Purpose

This version of Whisper Medium takes a recording of someone speaking Turkish and produces a written transcript. Training used the first 10% of the training and validation splits of Mozilla Common Voice 11.0. The model card reports a word error rate of 18.51% on its evaluation set, counting words replaced, missed or added against a reference transcript.

Base model or method
OpenAI Whisper Medium

Sources

Whisper Medium TR