app.py

import gradio as gr
from transformers import AutoProcessor, AutoModelForCTC
import torch
import soundfile as sf
import nemo.collections.asr as nemo_asr
import gradio as gr
# Load the model
model = nemo_asr.models.ASRModel.from_pretrained("nvidia/stt_en_fastconformer_hybrid_large_pc")

# Function to transcribe audio
def transcribe_audio(audio_file):
    transcription = model.transcribe([audio_file])
    return transcription[0]

# Gradio interface
iface = gr.Interface(
    fn=transcribe_audio,
    inputs=gr.Audio(type="filepath"),
    outputs="text",
    title="Real-Time Transcription with FastConformer"
)

iface.launch()