Add num2words

app.py

@@ -3,6 +3,9 @@ import tempfile
 from TTS.api import TTS
 from huggingface_hub import hf_hub_download
 import torch
+import json
+from snfl_imdann import TifinaghNumberConverter
+import re
 
 CUDA = torch.cuda.is_available()
 
@@ -32,8 +35,26 @@ my_examples = [
     ["ⵜⴻⵜⵜⵏ ⴰⴳ ⵡⵓⵛⵛⵏ, ⵜⵜⵔⵓⵏ ⵅ ⵓⵎⴽⵙⴰ.", "rif", "idj", False]
 ]
 
+best_model_path = hf_hub_download(repo_id=REPO_ID, filename="checkpoint_390000.pth") 
+config_path = hf_hub_download(repo_id=REPO_ID, filename="config.json")
+
+api = TTS(model_path=best_model_path, config_path=config_path).to("cuda" if CUDA else "cpu")
+
+# pre-download voice conversion models
+for model in VOICE_CONVERSION_MODELS.values():
+    api.load_vc_model_by_name(model, gpu=CUDA)
+
+with open(config_path, "r") as f:
+    config = json.load(f)
+    available_chars = config["characters"]["characters"]
+    available_punct = config["characters"]["punctuations"]
+
+available_chars = available_chars + "".join([str(i) for i in range(10)])
+
+placeholder = f"The available characters are: {available_chars} and the available punctuation is: {available_punct}"
+
 my_inputs = [
-    gr.Textbox(lines=5, label="Input Text", placeholder="The only available characters are: ⴰⴱⴳⴷⴹⴻⴼⴽⵀⵃⵄⵅⵇⵉⵊⵍⵎⵏⵓⵔⵕⵖⵙⵚⵛⵜⵟⵡⵢⵣⵥⵯ !,.:?"),
+    gr.Textbox(lines=5, label="Input Text", placeholder=placeholder),
     gr.Dropdown(label="Variant", choices=list(VARIANTS.items()), value="shi"),
     gr.Dropdown(label="Speaker", choices=SPEAKERS, value="yan"),
     gr.Checkbox(label="Split Sentences (each sentence will be generated separately)", value=False),
@@ -43,15 +64,6 @@ my_inputs = [
 
 my_outputs = gr.Audio(type="filepath", label="Output Audio", autoplay=True)
 
-best_model_path = hf_hub_download(repo_id=REPO_ID, filename="checkpoint_390000.pth") 
-config_path = hf_hub_download(repo_id=REPO_ID, filename="config.json")
-
-api = TTS(model_path=best_model_path, config_path=config_path).to("cuda" if CUDA else "cpu")
-
-# pre-download voice conversion models
-for model in VOICE_CONVERSION_MODELS.values():
-    api.load_vc_model_by_name(model, gpu=CUDA)
-
 def tts(text: str, variant: str = "shi", speaker: str = "yan", split_sentences: bool = False, speaker_wav: str = None, voice_cv_model: str = 'freevc24'):
     # replace oov characters
     text = text.replace("\n", ". ")
@@ -62,6 +74,9 @@ def tts(text: str, variant: str = "shi", speaker: str = "yan", split_sentences:
     text = text.replace(";", ",")
     text = text.replace("-", " ")
 
+    # convert numbers to their spoken form
+    text = re.sub(r"\d+", lambda x: TifinaghNumberConverter.convert(int(x.group(0))), text)
+
     with tempfile.NamedTemporaryFile(suffix = ".wav", delete = False) as fp:
         if speaker_wav:
             api.load_vc_model_by_name(VOICE_CONVERSION_MODELS[voice_cv_model], gpu=CUDA)
@@ -69,7 +84,7 @@ def tts(text: str, variant: str = "shi", speaker: str = "yan", split_sentences:
         else:
             api.tts_to_file(text, file_path=fp.name, split_sentences=split_sentences, speaker=speaker, language=variant)
 
-    return fp.name 
+    return fp.name
 
 iface = gr.Interface(
     fn=tts, 

snfl_imdann.py

@@ -0,0 +1,61 @@
+class TifinaghNumberConverter:
+    AND = " ⴷ "
+    UNITS = ["ⴰⵎⵢⴰ", "ⵢⴰⵏ", "ⵙⵉⵏ", "ⴽⵕⴰⴹ", "ⴽⴽⵓⵥ", "ⵙⵎⵎⵓⵙ", "ⵚⴹⵉⵚ", "ⵙⴰ", "ⵜⴰⵎ", "ⵜⵥⴰ"]
+    TENS = ["", "ⵎⵔⴰⵡ", "ⵙⵉⵎⵔⴰⵡ", "ⴽⵕⴰⵎⵔⴰⵡ", "ⴽⴽⵓⵎⵔⴰⵡ", "ⵙⵎⵎⵓⵎⵔⴰⵡ", "ⵚⴹⵉⵎⵔⴰⵡ", "ⵙⴰⵎⵔⴰⵡ", "ⵜⴰⵎⵔⴰⵡ", "ⵜⵥⴰⵎⵔⴰⵡ"]
+    HUNDREDS = ["", "ⵜⵉⵎⵉⴹⵉ", "ⵙⵏⴰⵜ ⵜⵎⴰⴹ", "ⴽⵕⴰⴹⵜ ⵜⵎⴰⴹ", "ⴽⴽⵓⵥⵜ ⵜⵎⴰⴹ", "ⵙⵎⵎⵓⵙⵜ ⵜⵎⴰⴹ", "ⵚⴹⵉⵚⵜ ⵜⵎⴰⴹ", "ⵙⴰⵜ ⵜⵎⴰⴹ", "ⵜⴰⵎⵜ ⵜⵎⴰⴹ", "ⵜⵥⴰⵜ ⵜⵎⴰⴹ"]
+    ORDERS = ["", "ⵉⴼⴹ", "ⴰⵎⵍⵢⵓⵏ", "ⴰⵎⵍⵢⴰⵕ", "ⴰⵜⵔⵉⵍⵢⵓⵏ", "ⴰⴽⵡⴰⴹⵕⵉⵍⵢⵓⵏ", "ⴰⴽⵡⵉⵏⵜⵔⵉⵍⵢⵓⵏ", "ⴰⵙⵉⴽⵙⵜⵉⵍⵢⵓⵏ", "ⴰⵙⵉⴱⵜⵉⵍⵢⵓⵏ", "ⴰⵡⴽⵜⵉⵍⵢⵓⵏ", "ⴰⵏⵓⵏⵉⵍⵢⵓⵏ"]
+    ORDERS_PLURAL = ["", "ⵡⴰⴼⴹⴰⵏ", "ⵉⴷ ⴰⵎⵍⵢⵓⵏ", "ⵉⴷ ⴰⵎⵍⵢⴰⵕ", "ⵉⵜⵔⵉⵍⵢⵓⵏⵏ", "ⵉⴽⵡⴰⴹⵕⵉⵍⵢⵓⵏⵏ", "ⵉⴽⵡⵉⵏⵜⵔⵉⵍⵢⵓⵏⵏ", "ⵉⵙⵉⴽⵙⵜⵉⵍⵢⵓⵏⵏ", "ⵉⵙⵉⴱⵜⵉⵍⵢⵓⵏⵏ", "ⵉⵡⴽⵜⵉⵍⵢⵓⵏⵏ", "ⵉⵏⵓⵏⵉⵍⵢⵓⵏⵏ"]
+
+    @classmethod
+    def convert_number_to_999(cls, n):
+        """Convert numbers from 0 to 999 to Tifinagh."""
+        if n == 0:
+            return ""
+        if n < 10:
+            return cls.UNITS[n]
+        if n < 20:
+            if n % 10 == 0:
+                return cls.TENS[n // 10]
+            return cls.UNITS[n % 10] + cls.AND + cls.TENS[n // 10]
+        if n < 100:
+            if n % 10 == 0:
+                return cls.TENS[n // 10]
+            return cls.TENS[n // 10] + cls.AND + cls.UNITS[n % 10]
+        if n < 1000:
+            if n % 100 == 0:
+                return cls.HUNDREDS[n // 100]
+            return cls.HUNDREDS[n // 100] + cls.AND + cls.convert_number_to_999(n % 100)
+        return ""
+
+    @classmethod
+    def convert_large_number(cls, n):
+        """Convert large numbers to Tifinagh."""
+        if n < 1000:
+            return cls.convert_number_to_999(n)
+        
+        for order in range(len(cls.ORDERS) - 1, 0, -1):
+            order_value = 10 ** (order * 3)
+            if n >= order_value:
+                quotient = n // order_value
+                remainder = n % order_value
+                
+                if quotient == 1:
+                    if remainder:
+                        return cls.ORDERS[order] + cls.AND + cls.convert_large_number(remainder)
+                    return cls.ORDERS[order]
+                else:
+                    if remainder:
+                        return cls.convert_number_to_999(quotient) + " " + \
+                               cls.ORDERS_PLURAL[order] + cls.AND + cls.convert_large_number(remainder)
+                    return cls.convert_number_to_999(quotient) + " " + cls.ORDERS_PLURAL[order]
+        
+        return ""
+
+    @classmethod
+    def convert(cls, number):
+        """Main conversion method."""
+        if number == 0:
+            return "ⴰⵎⵢⴰ"
+        if number < 0:
+            return "- " + cls.convert_large_number(abs(number))
+        return cls.convert_large_number(number)