change to process_text

infer_onnx.py

@@ -24,8 +24,15 @@ def intersperse(lst, item):
     return result
 
 
+def check_input(text):
+    if len(text) > 500:
+        raise gr.Error("Wrong input, more than 5 characters")
+    else:
+        return text
+
 def process_text(i: int, text: str, device: torch.device, cleaner:str):
     print(f"[{i}] - Input text: {text}")
+    check_input(text)
     x = torch.tensor(
         intersperse(text_to_sequence(text, [cleaner]), 0),
         dtype=torch.long,
@@ -36,12 +43,6 @@ def process_text(i: int, text: str, device: torch.device, cleaner:str):
     print(x_phones)
     return x.numpy(), x_lengths.numpy()
 
-def check_input(text):
-    if len(text) > 500:
-        raise gr.Error("Wrong input, more than 5 characters")
-    else:
-        return text
-
 # paths
 MODEL_PATH_MATCHA_MEL_BAL="matcha_multispeaker_cat_bal_opset_15_10_steps.onnx"
 MODEL_PATH_MATCHA_MEL_CAT="matcha_multispeaker_cat_cen_opset_15_10_steps.onnx"
@@ -162,7 +163,7 @@ def tts(text:str, accent:str, spk_name:str, temperature:float, length_scale:floa
     denoise=True
     spk_id = speaker_id_dict[accent][spk_name]
     sid = np.array([int(spk_id)]) if spk_id is not None else None
-    text_matcha , text_lengths = process_text(0,check_input(text),"cpu",cleaner=cleaners[accent])
+    text_matcha , text_lengths = process_text(0,text,"cpu",cleaner=cleaners[accent])
     model_matcha_mel = models[accent]
 
     # MATCHA VOCOS