Update app

app.py

@@ -17,19 +17,26 @@ gos_text2speech = Text2Speech.from_pretrained(
     noise_scale_dur=1.0
 )
 
-def inference(text,lang):
-  with torch.no_grad():
-      if lang == "Hoogelaandsters":
-          wav = gos_text2speech(text.lower(), sids=np.array([1]))["wav"]
-          sf.write("out.wav", wav.view(-1).cpu().numpy(), gos_text2speech.fs)
-      if lang == "Oldambsters":
-          wav = gos_text2speech(text.lower(), sids=np.array([2]))["wav"]
-          sf.write("out.wav", wav.view(-1).cpu().numpy(), gos_text2speech.fs)
-      if lang == "Westerkertaaiers":
-         wav = gos_text2speech(text.lower(), sids=np.array([3]))["wav"]
-         sf.write("out.wav", wav.view(-1).cpu().numpy(), gos_text2speech.fs)
-
-  return  "out.wav", "out.wav"
+def inference(text, lang):
+    with torch.no_grad():
+        lines = text.splitlines()
+        outputs = []
+
+        for line in lines:
+            line = line.lower()
+            if lang == "Hoogelaandsters":
+                wav = gos_text2speech(line, sids=np.array([1]))["wav"]
+            elif lang == "Oldambsters":
+                wav = gos_text2speech(line, sids=np.array([2]))["wav"]
+            elif lang == "Westerkertaaiers":
+                wav = gos_text2speech(line, sids=np.array([3]))["wav"]
+
+            outputs.append(wav)
+
+        concatenated_wav = np.concatenate([o.view(-1).cpu().numpy() for o in outputs])
+        sf.write("out.wav", concatenated_wav, gos_text2speech.fs)
+
+    return "out.wav", "out.wav"
 
 title = "Gronings text-to-speech"
 examples = [
@@ -39,7 +46,7 @@ examples = [
 gr.Interface(
     inference,
     [gr.inputs.Textbox(label="Input text", lines=3), gr.inputs.Radio(choices=["Hoogelaandsters", "Oldambsters", "Westerkertaaiers"], type="value", default="Hoogelaandsters", label="Variant")], 
-    [gr.outputs.Audio(type="file", label="Output"), gr.outputs.File()],
+    [gr.outputs.Audio(type="filepath", label="Output"), gr.outputs.File()],
     title=title,
     examples=examples
     ).launch(enable_queue=True)