Spaces:
Runtime error
Runtime error
Commit
·
969ab0b
1
Parent(s):
8cc4893
Update app.py
Browse files
app.py
CHANGED
|
@@ -1,6 +1,192 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
import streamlit as st
|
| 2 |
-
import
|
|
|
|
|
|
|
|
|
|
| 3 |
|
| 4 |
|
| 5 |
|
| 6 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
%%writefile app.py
|
| 2 |
+
from IPython.display import Javascript
|
| 3 |
+
from IPython import display
|
| 4 |
+
from google.colab import output
|
| 5 |
+
from base64 import b64decode
|
| 6 |
+
import datetime
|
| 7 |
+
import whisper
|
| 8 |
+
import openai
|
| 9 |
+
import os
|
| 10 |
+
import base64
|
| 11 |
+
from Crypto.Cipher import AES
|
| 12 |
+
from streamlit_bokeh_events import streamlit_bokeh_events
|
| 13 |
import streamlit as st
|
| 14 |
+
from bokeh.models.widgets import Button
|
| 15 |
+
from bokeh.models.widgets.buttons import Button
|
| 16 |
+
from bokeh.models import CustomJS
|
| 17 |
+
from streamlit_bokeh_events import streamlit_bokeh_events
|
| 18 |
|
| 19 |
|
| 20 |
|
| 21 |
+
RECORD = """
|
| 22 |
+
const sleep = time => new Promise(resolve => setTimeout(resolve, time))
|
| 23 |
+
const b2text = blob => new Promise(resolve => {
|
| 24 |
+
const reader = new FileReader()
|
| 25 |
+
reader.onloadend = e => resolve(e.srcElement.result)
|
| 26 |
+
reader.readAsDataURL(blob)
|
| 27 |
+
})
|
| 28 |
+
var record = time => new Promise(async resolve => {
|
| 29 |
+
stream = await navigator.mediaDevices.getUserMedia({ audio: true })
|
| 30 |
+
recorder = new MediaRecorder(stream)
|
| 31 |
+
chunks = []
|
| 32 |
+
recorder.ondataavailable = e => chunks.push(e.data)
|
| 33 |
+
recorder.start()
|
| 34 |
+
await sleep(time)
|
| 35 |
+
recorder.onstop = async ()=>{
|
| 36 |
+
blob = new Blob(chunks)
|
| 37 |
+
text = await b2text(blob)
|
| 38 |
+
resolve(text)
|
| 39 |
+
}
|
| 40 |
+
recorder.stop()
|
| 41 |
+
})
|
| 42 |
+
"""
|
| 43 |
+
|
| 44 |
+
openai.api_key = os.environ["API_KEY"]
|
| 45 |
+
|
| 46 |
+
with open("encrypt.txt", "r") as encfile:
|
| 47 |
+
encoder_txt = encfile.read()
|
| 48 |
+
|
| 49 |
+
with open("decrypt.txt", "r") as decfile:
|
| 50 |
+
decoder_txt = decfile.read()
|
| 51 |
+
|
| 52 |
+
def openai_fun(myprompt):
|
| 53 |
+
response_encoded = openai.Completion.create(
|
| 54 |
+
engine="text-davinci-003",
|
| 55 |
+
prompt = myprompt,
|
| 56 |
+
max_tokens=1024,
|
| 57 |
+
n=1,
|
| 58 |
+
stop=None,
|
| 59 |
+
temperature=0.5,
|
| 60 |
+
)
|
| 61 |
+
return response_encoded
|
| 62 |
+
|
| 63 |
+
def record(sec=5):
|
| 64 |
+
display.display(Javascript(RECORD))
|
| 65 |
+
s = output.eval_js('record(%d)' % (sec*1000))
|
| 66 |
+
b = b64decode(s.split(',')[1])
|
| 67 |
+
ts = datetime.datetime.now()
|
| 68 |
+
filename = ts.strftime("%Y_%m_%d_%H_%M_%S")
|
| 69 |
+
with open(f'{filename}.wav','wb') as f:
|
| 70 |
+
f.write(b)
|
| 71 |
+
return f'{filename}.wav' # or webm ?
|
| 72 |
+
|
| 73 |
+
model = whisper.load_model("base")
|
| 74 |
+
transcribed = []
|
| 75 |
+
|
| 76 |
+
while True:
|
| 77 |
+
user_choice = st.text_input("Do you want to record a new audio for transcription?[y/n]")
|
| 78 |
+
|
| 79 |
+
if user_choice == 'y':
|
| 80 |
+
st.write('Recording! (5 seconds)')
|
| 81 |
+
record(5)
|
| 82 |
+
folder_path = "/content"
|
| 83 |
+
audio_files = [f for f in os.listdir(folder_path) if f.endswith(".wav")]
|
| 84 |
+
|
| 85 |
+
audio_files.sort(key=lambda x: os.path.getmtime(os.path.join(folder_path, x)), reverse=True)
|
| 86 |
+
last_audio_file_path = os.path.join(folder_path, audio_files[0])
|
| 87 |
+
st.write('Transcribing audio file: ',last_audio_file_path)
|
| 88 |
+
# COMMENT IF NOT NEEDED:
|
| 89 |
+
if os.path.exists(last_audio_file_path) and not last_audio_file_path in transcribed:
|
| 90 |
+
audio = whisper.load_audio(last_audio_file_path)
|
| 91 |
+
audio = whisper.pad_or_trim(audio)
|
| 92 |
+
mel = whisper.log_mel_spectrogram(audio).to(model.device)
|
| 93 |
+
options = whisper.DecodingOptions(language= 'en', fp16=False)
|
| 94 |
+
|
| 95 |
+
result = whisper.decode(model, mel, options)
|
| 96 |
+
|
| 97 |
+
if result.no_speech_prob < 0.5:
|
| 98 |
+
mymsg = result.text
|
| 99 |
+
st.write("Actual Message: ",mymsg)
|
| 100 |
+
|
| 101 |
+
enc_prompt = encoder_txt + mymsg
|
| 102 |
+
|
| 103 |
+
openai_fun(enc_prompt)
|
| 104 |
+
|
| 105 |
+
if openai_fun(enc_prompt)['choices'][0]['text'] != "":
|
| 106 |
+
# print(response_encoded['choices'][0]['text'])
|
| 107 |
+
exec(openai_fun(enc_prompt)['choices'][0]['text'])
|
| 108 |
+
encoded_msg = enc(mymsg)
|
| 109 |
+
print("The encoded message: ", encoded_msg)
|
| 110 |
+
|
| 111 |
+
decode_ = st.text_input("Do you wish to decode the message?[y/n]")
|
| 112 |
+
if decode_ == "y":
|
| 113 |
+
dec_prompt = decoder_txt + str(encoded_msg)
|
| 114 |
+
response_decoded = openai.Completion.create(
|
| 115 |
+
engine="text-davinci-003",
|
| 116 |
+
prompt = dec_prompt,
|
| 117 |
+
max_tokens=500,
|
| 118 |
+
n=1,
|
| 119 |
+
stop=None,
|
| 120 |
+
temperature=0.5,
|
| 121 |
+
)
|
| 122 |
+
|
| 123 |
+
if response_decoded['choices'][0]['text'] != "":
|
| 124 |
+
print(response_decoded['choices'][0]['text'])
|
| 125 |
+
exec(response_decoded['choices'][0]['text'])
|
| 126 |
+
decoded_msg = dec(encoded_msg, key)
|
| 127 |
+
print("The decoded message: ", decoded_msg)
|
| 128 |
+
|
| 129 |
+
|
| 130 |
+
else:
|
| 131 |
+
st.write('Retry! The message could')
|
| 132 |
+
|
| 133 |
+
|
| 134 |
+
break # exit the loop
|
| 135 |
+
|
| 136 |
+
elif user_choice == 'n':
|
| 137 |
+
uc1 = input('Do you want to transcribe an existing audio?[y/n]')
|
| 138 |
+
|
| 139 |
+
if uc1 == 'y':
|
| 140 |
+
folder_path = "/content"
|
| 141 |
+
audio_files = [f for f in os.listdir(folder_path) if f.endswith(".wav")]
|
| 142 |
+
print('Audio files present: ',audio_files)
|
| 143 |
+
|
| 144 |
+
audio_files.sort(key=lambda x: os.path.getmtime(os.path.join(folder_path, x)), reverse=True)
|
| 145 |
+
last_audio_file_path = os.path.join(folder_path, audio_files[0])
|
| 146 |
+
print('Transcribing last audio file: ',last_audio_file_path)
|
| 147 |
+
|
| 148 |
+
# COMMENT IF NOT NEEDED:
|
| 149 |
+
if os.path.exists(last_audio_file_path) and not last_audio_file_path in transcribed:
|
| 150 |
+
audio = whisper.load_audio(last_audio_file_path)
|
| 151 |
+
audio = whisper.pad_or_trim(audio)
|
| 152 |
+
mel = whisper.log_mel_spectrogram(audio).to(model.device)
|
| 153 |
+
options = whisper.DecodingOptions(language= 'en', fp16=False)
|
| 154 |
+
|
| 155 |
+
result = whisper.decode(model, mel, options)
|
| 156 |
+
|
| 157 |
+
if result.no_speech_prob < 0.5:
|
| 158 |
+
mymsg = result.text
|
| 159 |
+
print("Actual Message: ",mymsg)
|
| 160 |
+
enc_prompt = encoder_txt + result.text
|
| 161 |
+
response_encoded = openai.Completion.create(
|
| 162 |
+
engine="text-davinci-003",
|
| 163 |
+
prompt = enc_prompt,
|
| 164 |
+
max_tokens=1024,
|
| 165 |
+
n=1,
|
| 166 |
+
stop=None,
|
| 167 |
+
temperature=0.5,
|
| 168 |
+
)
|
| 169 |
+
|
| 170 |
+
if response_encoded['choices'][0]['text'] != "":
|
| 171 |
+
# print(response_encoded['choices'][0]['text'])
|
| 172 |
+
exec(response_encoded['choices'][0]['text'])
|
| 173 |
+
encoded_msg = enc(mymsg)
|
| 174 |
+
st.write("The encoded message: ", encoded_msg)
|
| 175 |
+
|
| 176 |
+
else:
|
| 177 |
+
st.write('Retry! The message could')
|
| 178 |
+
|
| 179 |
+
# DELETE audio
|
| 180 |
+
|
| 181 |
+
break # exit the loop
|
| 182 |
+
|
| 183 |
+
elif uc1 == 'n':
|
| 184 |
+
continue # continue the loop, prompting for input again
|
| 185 |
+
|
| 186 |
+
else:
|
| 187 |
+
st.write('Invalid input, please enter y or n')
|
| 188 |
+
continue # continue the loop, prompting for input again
|
| 189 |
+
|
| 190 |
+
else:
|
| 191 |
+
st.write('Invalid input, please enter y or n')
|
| 192 |
+
continue # continue the loop, prompting for input again
|