app.py

# >>>>>>   Adapted/frankensteined from these scripts:   <<<<<<<
#                 for Summary Interface:
# >>>> https://huggingface.co/spaces/khxu/pegasus-text-summarizers/blob/main/app.py
#                      Audio Interface
# >>>> https://huggingface.co/spaces/iSky/Speech-audio-to-text-with-grammar-correction/blob/main/app.py
#                      Gramar
# >>>> https://huggingface.co/deep-learning-analytics/GrammarCorrector/blob/main/README.md


import gradio as gr
from transformers import pipeline
from gradio.mix import Parallel, Series

# >>>>>>>>>>>>>>>>>>>> Danger Below <<<<<<<<<<<<<<<<<<<<<<
# Load Interfaces:
s2t = gr.Interface.load('huggingface/facebook/s2t-medium-librispeech-asr')
grammar = gr.Interface.load('huggingface/deep-learning-analytics/GrammarCorrector')
sum_it = gr.Interface.load('huggingface/SamuelMiller/lil_sum_sum') 

# Audio Functions:
def out(audio):
  flag = True
  if audio==None:
    return "no audio" 
  
  else:               
    a = s2t(audio)
    #g = grammar(a)
    #s = sum_it(g)                              # Summarize Audio with sum_it
    return a, a #grammar(a, num_return_sequences=1) # grammar(s),                       # Grammar Filter 

  #else:
   # return "something is wrong in the function?"

# Construct Interfaces:
iface = gr.Interface(
  fn=out, 
  title="Speech Audio to text (with corrected grammar)",
  description="Let's Hear It!! This app transforms your speech (input) to text with corrected grammar after (output)!",
  inputs= gr.inputs.Audio(source="microphone", type="filepath", label=None, optional=True),
  outputs= ['text','text']
)

# Launch Interface
iface.launch(enable_queue=True,show_error=True)

  # From Original Code:
# gr.inputs.Audio(source="upload", type="filepath", label=None, optional=True),  
# examples=[["Grammar-Correct-Sample.mp3"], ["Grammar-Wrong-Sample.mp3"],],

#def speech_to_text(inp):
    #pass  # speech recognition model defined here

#gr.Interface(speech_to_text, inputs="mic", outputs=gr.Textbox(label="Predicted text", lines=4))