app.py

import torch
import gradio as gr
import pytube as pt
from transformers import pipeline
from huggingface_hub import model_info
import time
import unicodedata

MODEL_NAME = "SakshiRathi77/wav2vec2-large-xlsr-300m-hi-kagglex"
lang = "hi"

device = 0 if torch.cuda.is_available() else "cpu"
pipe = pipeline(
    task="automatic-speech-recognition",
    model=MODEL_NAME,
    device=device,
)

def transcribe(microphone, file_upload):
    warn_output = ""
    if (microphone is not None) and (file_upload is not None):
        warn_output = (
            "WARNING: You've uploaded an audio file and used the microphone. "
            "The recorded file from the microphone will be used and the uploaded audio will be discarded.\n"
        )

    elif (microphone is None) and (file_upload is None):
        return "ERROR: You have to either use the microphone or upload an audio file"
    file = microphone if microphone is not None else file_upload
    text = pipe(file)["text"]
    
    return warn_output + text


def rt_transcribe(audio, state=""):
    time.sleep(2)
    text = pipe(audio)["text"]
    state += unicodedata.normalize("NFC",text) + " "
    return state, state

demo = gr.Blocks()
examples=[["examples/example1.mp3"], ["examples/example2.mp3"],["examples/example3.mp3"]]




description = """
<p style="font-family: 'Arial', sans-serif; font-size: 24px; font-weight: bold;">
<center>
Welcome to the HindiSpeechPro, a cutting-edge interface powered by a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common_voice dataset. Easily convert your spoken words to accurate text with just a few clicks.
<img src="https://huggingface.co/spaces/SakshiRathi77/SakshiRathi77-Wav2Vec2-hi-kagglex/resolve/main/Images/Hindi-Speech-Voice-Recognition-Tool.jpg" alt="logo" ;>
</center>
</p>
</div>
"""


mf_transcribe = gr.Interface(
    fn=transcribe,
    inputs=[
        gr.inputs.Audio(source="microphone", type="filepath"),
        gr.inputs.Audio(source="upload", type="filepath"),
    ],
    outputs="text",
    theme="huggingface",
    title="HindiSpeechPro: WAV2VEC-Powered ASR Interface",
    description= description ,
    allow_flagging="never",
    examples=examples,
    css=css,
)

rt_transcribe = gr.Interface(
    fn=rt_transcribe,
    inputs=[
        gr.Audio(source="microphone", type="filepath", streaming=True), 
        "state"
    ],
    outputs=[ "textbox",
        "state"],
    theme="huggingface",
    title="HindiSpeechPro: WAV2VEC-Powered ASR Interface",
    description= description ,
    allow_flagging="never",
    live=True,
)


with demo:
    gr.TabbedInterface([mf_transcribe, rt_transcribe], ["Transcribe Audio", "Transcribe Realtime Voice"])

demo.launch(share=True)