Перевод аудио в текст с помощью нейронной сети📌 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