Update app.py

app.py

@@ -5,33 +5,18 @@ import torch
 import pyewts
 import noisereduce as nr
 from transformers import SpeechT5Processor, SpeechT5ForTextToSpeech, SpeechT5HifiGan
-def remove_repeated_words(text):
-    # Tokenize the input text into words
-    words = text.split()
+from num2tib.core import convert
+from num2tib.core import convert2text
+import re
 
-    # Create a dictionary to count word occurrences
-    word_count = {}
-
-    # Create a list to store the final words
-    new_words = []
-
-    for word in words:
-        # Check if the word is in the dictionary
-        if word in word_count:
-            # If it has occurred once before, add it to the list with a count of 2
-            if word_count[word] == 1:
-                new_words.append(word)
-                word_count[word] = 2
-        else:
-            # If it has not occurred before, add it to the dictionary with a count of 1
-            word_count[word] = 1
-            new_words.append(word)
-
-    # Join the modified words back into a string
-    result = ' '.join(new_words)
+def replace_numbers_with_convert(sentence, wylie=True):
+    pattern = r'\d+'
+    def replace(match):
+        return convert(match.group(), wylie)
+    result = re.sub(pattern, replace, sentence)
+    
     return result
 
-
 converter = pyewts.pyewts()
 checkpoint = "TenzinGayche/TTS_run3_ep20_174k_b"
 processor = SpeechT5Processor.from_pretrained(checkpoint)
@@ -65,6 +50,7 @@ def predict(text, speaker):
         return (16000, np.zeros(0).astype(np.int16))
     text = converter.toWylie(text)
     text=cleanup_text(text)
+    text=replace_numbers_with_convert(text)
     inputs = processor(text=text, return_tensors="pt")
     # limit input length
     input_ids = inputs["input_ids"]