app.py

from random import sample
import gradio as gr
import torchaudio
import torch
import torch.nn as nn
import lightning_module
import pdb
import jiwer
from local.convert_metrics import nat2avaMOS, WER2INTELI
from local.indicator_plot import Intelligibility_Plot, Naturalness_Plot
from local.pitch_contour import draw_spec_db_pitch
# ASR part
from transformers import pipeline
p = pipeline("automatic-speech-recognition")

# WER part
transformation = jiwer.Compose([
    jiwer.ToLowerCase(),
    jiwer.RemoveWhiteSpace(replace_by_space=True),
    jiwer.RemoveMultipleSpaces(),
    jiwer.ReduceToListOfListOfWords(word_delimiter=" ")
])

# WPM part
from transformers import Wav2Vec2Processor, Wav2Vec2ForCTC
# processor = Wav2Vec2Processor.from_pretrained("facebook/wav2vec2-xlsr-53-espeak-cv-ft")
# phoneme_model = Wav2Vec2ForCTC.from_pretrained("facebook/wav2vec2-xlsr-53-espeak-cv-ft")
# phoneme_model =  pipeline(model="facebook/wav2vec2-xlsr-53-espeak-cv-ft") 
class ChangeSampleRate(nn.Module):
    def __init__(self, input_rate: int, output_rate: int):
        super().__init__()
        self.output_rate = output_rate
        self.input_rate = input_rate

    def forward(self, wav: torch.tensor) -> torch.tensor:
        # Only accepts 1-channel waveform input
        wav = wav.view(wav.size(0), -1)
        new_length = wav.size(-1) * self.output_rate // self.input_rate
        indices = (torch.arange(new_length) * (self.input_rate / self.output_rate))
        round_down = wav[:, indices.long()]
        round_up = wav[:, (indices.long() + 1).clamp(max=wav.size(-1) - 1)]
        output = round_down * (1. - indices.fmod(1.)).unsqueeze(0) + round_up * indices.fmod(1.).unsqueeze(0)
        return output

model = lightning_module.BaselineLightningModule.load_from_checkpoint("epoch=3-step=7459.ckpt").eval()

def calc_mos(audio_path, ref):
    wav, sr = torchaudio.load(audio_path, channels_first=True)  
    if wav.shape[0] > 1:
        wav = wav.mean(dim=0, keepdim=True) # Mono channel
    # get decibel
    
    osr = 16_000
    batch = wav.unsqueeze(0).repeat(10, 1, 1)
    csr = ChangeSampleRate(sr, osr)
    out_wavs = csr(wav)
    db = torchaudio.transforms.AmplitudeToDB(stype="amplitude", top_db=80)(wav)
    # ASR
    trans = p(audio_path)["text"]
    # WER
    wer = jiwer.wer(ref, trans, truth_transform=transformation, hypothesis_transform=transformation)
    # WER convert to Intellibility score
    INTELI_score = WER2INTELI(wer*100)
    
    INT_fig = Intelligibility_Plot(INTELI_score)
    
    # MOS
    batch = {
        'wav': out_wavs,
        'domains': torch.tensor([0]),
        'judge_id': torch.tensor([288])
    }
    with torch.no_grad():
        output = model(batch)
    predic_mos = output.mean(dim=1).squeeze().detach().numpy()*2 + 3
    # MOS to AVA MOS
    AVA_MOS = nat2avaMOS(predic_mos)
    
    MOS_fig = Naturalness_Plot(AVA_MOS)
    
    # Phonemes per minute (PPM)
    # with torch.no_grad():
    #     logits = phoneme_model(out_wavs).logits
    # phone_predicted_ids = torch.argmax(logits, dim=-1)
    # phone_transcription = processor.batch_decode(phone_predicted_ids)
    
    # Disable PPM for now
    phone_transcription = ['D U M M Y']
    lst_phonemes = phone_transcription[0].split(" ")
    wav_vad = torchaudio.functional.vad(wav, sample_rate=sr)

    # draw f0 and db analysis plot
    f0_db_fig = draw_spec_db_pitch(audio_path, save_fig_path=None)

    ppm = len(lst_phonemes) / (wav_vad.shape[-1] / sr) * 60
    
    # pdb.set_trace()
    # return AVA_MOS, MOS_fig, INTELI_score, INT_fig, trans, phone_transcription, ppm, f0_db_fig
    return AVA_MOS, MOS_fig, INTELI_score, INT_fig, trans, f0_db_fig


with open("local/description.md") as f:
    description = f.read()

# x = calc_mos("JOHN1.wav", "he would answer in a soft voice, 'I don't know.'")
# pdb.set_trace()

examples = [
    ["local/Julianna_Set1_Author_01.wav", "Once upon a time, there was a young rat named Arthur who couldn't make up his mind."],
    ["local/Patient_Arthur_set1_002_noisy.wav", "Whenever the other rats asked him if he would like to go hunting with them, he would answer in a soft voice, 'I don't know.'"],
]
iface = gr.Interface(
  fn=calc_mos,
  inputs=[gr.Audio(type='filepath', label="Audio to evaluate"), 
          gr.Textbox(placeholder="Input reference here (Don't keep this empty)", label="Reference")],
  outputs=[gr.Textbox(placeholder="Naturalness Score, ranged from 1 to 5, the higher the better.", label="Naturalness Score, ranged from 0 to 5, the higher the better.", visible=False),
           gr.Plot(label="Naturalness Score, ranged from 1 to 5, the higher the better.", show_label=True, container=True),
           gr.Textbox(placeholder="Intelligibility Score", label = "Intelligibility Score, range from 0 to 100, the higher the better", visible=False),
           gr.Plot(label="Intelligibility Score, range from 0 to 100, the higher the better", show_label=True, container=True),
           gr.Textbox(placeholder="Hypothesis", label="Hypothesis"),
           gr.Plot(label="Pitch Contour and dB Analysis", show_label=True, container=True)],
  title="Speech Analysis by Laronix AI",
  description=description,
  allow_flagging="auto",
  examples=examples,
)   
        # Currently remove PPM and Phonemes
        #    gr.Textbox(placeholder="Predicted Phonemes", label="Predicted Phonemes", visible=False),
        #    gr.Textbox(placeholder="Speaking Rate, Phonemes per minutes", label="Speaking Rate, Phonemes per minutes", visible=False),
# add password to protect the interface
iface.launch(share=False, auth=['Laronix', 'LaronixSLP'], auth_message="Authentication Required, ask kevin@laronix.com for password.\n Thanks for your cooperation!")