app.py

import torch
#from transformers import pipeline
#from transformers.pipelines.audio_utils import ffmpeg_read
from speechscore import SpeechScore 
import gradio as gr

MODEL_NAME = "alibabasglab/speechscore"
BATCH_SIZE = 1

device = 0 if torch.cuda.is_available() else "cpu"

mySpeechScore = SpeechScore([
        'SRMR'
    ])


# Copied from https://github.com/openai/whisper/blob/c09a7ae299c4c34c5839a76380ae407e7d785914/whisper/utils.py#L50
def format_timestamp(seconds: float, always_include_hours: bool = False, decimal_marker: str = "."):
    if seconds is not None:
        milliseconds = round(seconds * 1000.0)

        hours = milliseconds // 3_600_000
        milliseconds -= hours * 3_600_000

        minutes = milliseconds // 60_000
        milliseconds -= minutes * 60_000

        seconds = milliseconds // 1_000
        milliseconds -= seconds * 1_000

        hours_marker = f"{hours:02d}:" if always_include_hours or hours > 0 else ""
        return f"{hours_marker}{minutes:02d}:{seconds:02d}{decimal_marker}{milliseconds:03d}"
    else:
        # we have a malformed timestamp so just return it as is
        return seconds


def score(file, task, return_timestamps):
    scores = mySpeechScore(test_path=file, reference_path=None, window=None, score_rate=16000, return_mean=True)
    return scores


demo = gr.Blocks()

mic_score = gr.Interface(
    fn=score,
    inputs=[
        gr.Audio(sources=["microphone"],
                waveform_options=gr.WaveformOptions(
                waveform_color="#01C6FF",
                waveform_progress_color="#0066B4",
                skip_length=2,
                show_controls=False,
                ),
            ),
        gr.Radio(["absolute_score", "relative_score"], label="Task", default="absolute_score"),
        gr.Checkbox(default=False, label="Return timestamps"),
    ],
    outputs="text",
    layout="horizontal",
    theme="huggingface",
    title="Score speech from microphone",
    description=(
        "Score audio inputs with the click of a button! Demo uses the"
        " commonly used speech quality assessment methods for the audio files"
        " of arbitrary length."
    ),
    allow_flagging="never",
)

file_score = gr.Interface(
    fn=score,
    inputs=[
        gr.Audio(sources=["upload"], optional=True, label="Audio file", type="filepath"),
        gr.Radio(["absolute_score", "relative_score"], label="Task", default="absolute_score"),
        gr.Checkbox(default=False, label="Return timestamps"),
    ],
    outputs="text",
    layout="horizontal",
    theme="huggingface",
    title="Score speech from a file",
    description=(
        "Score audio inputs with the click of a button! Demo uses the"
        " commonly used speech quality assessment methods for the audio files"
        " of arbitrary length."
    ),
    examples=[
        ["./example.flac", "score", False],
        ["./example.flac", "score", True],
    ],
    cache_examples=True,
    allow_flagging="never",
)

with demo:
    gr.TabbedInterface([mic_score, file_score], ["Score Microphone", "Score Audio File"])

demo.launch(enable_queue=True)