Load and process audio file to standard sample rate using librosa

Translate audio to text using Whisper

Calculate WER based on transcribed texts using jiwer

Calculate PESQ based on librosa processed audio using pesq

Calculate MOS based on librosa processed audio using NISQA

Load Original Audio (librosa)

Resample Original Audio (librosa)

Load Translated Audio (librosa)

Resample Translated Audio (librosa)

Load Original Audio (Whisper)

Transcribe Original Audio (Whisper)

Load Translated Audio (Whisper)

Transcribe Translated Audio (Whisper)

Load Original Audio (Whisper)

Transcribe Original Audio (Whisper)

Encode Transcribed Original Text

Load Translated Audio (Whisper)

Transcribe Translated Audio (Whisper)

Encode Transcribed Translated Text

Load Original Audio (librosa)

Resample Original Audio (librosa)

Encode Resampled Original Audio

Load Translated Audio (librosa)

Resample Translated Audio (librosa)

Encode Resampled Translated Audio

Compute PESQ (pesq)

PESQ Result Range 1-4.5

Compute MOS (NISQA)

MOS Result Range 1-5

Compute STOI (pystoi)

STOI Result Range 0-1

Compute SNR (r_calc)

SNR Result Range 0-100

Compute WER (jiwer)

WER Result Range 0-1

Compute Cosine Similarity (Text Embedding)

Compute Cosine Similarity (Audio Embedding)

Cosine Similarity Result Range 0-1

Original Audio

Translated Audio