|
import gradio as gr |
|
from transformers import pipeline |
|
from pydub import AudioSegment |
|
|
|
pipe = pipeline("automatic-speech-recognition", "facebook/wav2vec2-base-960h") |
|
|
|
dictionary = {' ': '/', |
|
"'": '.----.', |
|
'0': '-----', |
|
'1': '.----', |
|
'2': '..---', |
|
'3': '...--', |
|
'4': '....-', |
|
'5': '.....', |
|
'6': '-....', |
|
'7': '--...', |
|
'8': '---..', |
|
'9': '----.', |
|
'?': '..--..', |
|
'A': '.-', |
|
'B': '-...', |
|
'C': '-.-.', |
|
'D': '-..', |
|
'E': '.', |
|
'F': '..-.', |
|
'G': '--.', |
|
'H': '....', |
|
'I': '..', |
|
'J': '.---', |
|
'K': '-.-', |
|
'L': '.-..', |
|
'M': '--', |
|
'N': '-.', |
|
'O': '---', |
|
'P': '.--.', |
|
'Q': '--.-', |
|
'R': '.-.', |
|
'S': '...', |
|
'T': '-', |
|
'U': '..-', |
|
'V': '...-', |
|
'W': '.--', |
|
'X': '-..-', |
|
'Y': '-.--', |
|
'Z': '--..', |
|
} |
|
|
|
|
|
|
|
def asr(speech): |
|
dummy = AudioSegment.from_file("dummy.mp3", format="mp3") |
|
dot = AudioSegment.from_file("dot.mp3", format="mp3") |
|
dash = AudioSegment.from_file("dash.mp3", format="mp3") |
|
space = AudioSegment.from_file("space.mp3", format="mp3") |
|
i=0 |
|
transcript = pipe(speech)['text'] |
|
morseCode = "" |
|
for character in transcript: |
|
morseCode += dictionary[character] + " " |
|
while i<len(morseCode): |
|
if morseCode[i] == ".": |
|
dummy = dummy + dot |
|
i = i+1 |
|
elif morseCode[i] == "-": |
|
dummy = dummy + dash |
|
i = i+1 |
|
elif morseCode[i] == " ": |
|
dummy = dummy + space |
|
i = i+1 |
|
else: |
|
i = i+1 |
|
morseAudio = dummy.export("output.mp3", format="mp3") |
|
return transcript, morseCode, "output.mp3" |
|
|
|
gr.Interface(fn=asr, |
|
inputs = gr.inputs.Audio(source="microphone", type="filepath", optional=False, label="Please record your voice"), |
|
|
|
outputs = [gr.outputs.Textbox(type="str", label="Text Translation"), |
|
gr.outputs.Textbox(type="str", label="Morse Code"), |
|
gr.outputs.Audio(type="file", label="Audio Translation")], |
|
title = "Speech to Morse Code", |
|
description = "This app will translate your speech into text, morse code, and audio using wav2vec2-base-960h.", |
|
examples = [["helloworld.mp3"], ["SOS.mp3"], ["SMS.mp3"]], |
|
article = "Model: <a href=\"https://huggingface.co/facebook/wav2vec2-base-960h\">Wav2Vec2-Base-960h</a>" |
|
).launch() |
