SpeechScore / app.py
alibabasglab's picture
Update app.py
0b97e6d verified
import torch
from speechscore import SpeechScore
import gradio as gr
import pprint
def score_file(test_file, ref_file, score_list_nis, score_list_is):
score_list = []
if score_list_nis is not None:
score_list += score_list_nis
if score_list_is is not None:
score_list += score_list_is
mySpeechScore = SpeechScore(score_list)
scores = mySpeechScore(test_path=test_file, reference_path=ref_file, window=None, score_rate=16000, return_mean=False, round_digits=2)
return scores
def score_mic(test_file, score_list_nis):
score_list = []
if score_list_nis is not None:
score_list += score_list_nis
mySpeechScore = SpeechScore(score_list)
scores = mySpeechScore(test_path=test_file, reference_path=None, window=None, score_rate=16000, return_mean=False, round_digits=2)
return scores
demo = gr.Blocks()
file_score = gr.Interface(
fn=score_file,
inputs=[
gr.Audio(sources=["upload"], label="test file", type="filepath"),
gr.Audio(sources=["upload"], label="reference file", type="filepath"),
gr.Dropdown(
["DNSMOS", "SRMR"], value=["DNSMOS"], multiselect=True, label="Non-Intrusive Scores", info="Choose scores to include, reference audio is not required."
),
gr.Dropdown(
["PESQ", 'NB_PESQ', 'STOI', 'SISDR',
'FWSEGSNR', 'LSD', 'BSSEval',
'SNR', 'SSNR', 'LLR', 'CSIG', 'CBAK',
'COVL', 'MCD'], value=["PESQ", "STOI"], multiselect=True, label="Intrusive Scores", info="Choose scores to include, reference audio is required."
),
],
outputs="text",
title="Test your speech quality using <a href='https://github.com/modelscope/ClearerVoice-Studio/tree/main/speechscore' target='_blank'>speechscore</a></p>",
description=(
"Score speech quality with the click of a button! Demo includes the"
" commonly used speech quality assessments for the audio file"
" of arbitrary length."
),
)
mic_score = gr.Interface(
fn=score_mic,
inputs=[
gr.Audio(sources=["microphone"],
waveform_options=gr.WaveformOptions(
waveform_color="#01C6FF",
waveform_progress_color="#0066B4",
skip_length=2,
show_controls=False,
),
type='numpy',
),
gr.Dropdown(
["DNSMOS", "SRMR"], value=["DNSMOS"], multiselect=True, label="Non-Intrusive Scores", info="Choose scores to include, no reference audio is required."
),
],
outputs="text",
title="Test your microphone quality using <a href='https://github.com/modelscope/ClearerVoice-Studio/tree/main/speechscore' target='_blank'>speechscore</a></p>",
description=(
"Score your microphone quality with the click of a button!"
" Uses the most popular method to test your microphone quality"
" with a short speech clip."
),
#article = ("<a href='https://github.com/modelscope/ClearerVoice-Studio/tree/main/speechscore' target='_blank'>Github Repo</a></p> for SpeechScore")
)
with demo:
gr.TabbedInterface([mic_score, file_score], ["Score Microphone Quality", "Score Speech Quality"])
demo.launch()