Часть 2
audio = whisper.pad_or_trim(audio_13)
#Number of samples in our trimmed/padded audio
n_samples = audio.shape[-1]
#Time of each sample
time = np.linspace(0,(n_samples-1)*delta,n_samples)
plt.figure(figsize=(20,10))
plt.title('Signal')
plt.plot(time,audio)
plt.ylabel('amplitude')
plt.xlabel('seconds')
plt.show()
"""Next, we can start plotting a mel spectogram by applying a log_mel_spectogram() funtion to our audio file. It converts the y-axis (frequency) into the mel scale:"""
mel = whisper.log_mel_spectrogram(audio).to(model_m.device)
fig, (ax1, ax2) = plt.subplots(2)
fig.tight_layout(pad=5.0)
ax1.plot(time,audio)
ax1.set_title('Signal')
ax1.set_xlabel('Time, seconds')
ax1.set_ylabel('Amplitude')
ax2.imshow((mel.numpy()*mel.numpy())**(1/2),interpolation='nearest', aspect='auto')
ax2.set_title('Mel Spectrogram of a Signal')
ax2.set_xlabel('Time, seconds')
ax2.set_ylabel('Mel Scale')
#Next, we can move on to language detection.
#Language detection
sr=22050
ipd.Audio(audio, rate=sr)
probs = model_m.detect_language(mel)
probs
#Transcription
file_path2= 'Rec.mp3'
transcription = model_m.transcribe(file_path2, fp16 = False)['text']
transcription
