Edit model card

Voice Transcript - Portuguese Focused (From Whisper)

Optimized to work with Brazilian Language.

Current Version: v0.1

Downloads last month: 73

Safetensors

Model size

764M params

Tensor type

F32

·

Dataset used to train cloudqi/cqi_speech_recognize_pt_v0

Spaces using cloudqi/cqi_speech_recognize_pt_v0 2

Evaluation results

Test WER on LibriSpeech (clean)
test set self-reported

2.900
Test WER on LibriSpeech (other)
test set self-reported

5.900
Test WER on Common Voice 11.0
test set self-reported

53.870

View on Papers With Code