Add support for African voices.

app.py

@@ -63,6 +63,7 @@ type=['wav'])
                 finetuned_mms4 = synth_mms(tts_text, "khof312/mms-tts-spa-female")
             if tts_lang=="lin":
                 finetuned_mms1 = synth_mms(tts_text, "khof312/mms-tts-lin-female")
+                finetuned_africanvoices = synth_africanvoices(tts_text, models[tts_lang]['africanvoices'])
             
             #vc_mms
             #vc_coqui
@@ -152,14 +153,20 @@ type=['wav'])
                 "### Fine Tuned" 
                 row1 = st.columns([1,1,2])
                 row2 = st.columns([1,1,2])
+                row3 = st.columns([1,1,2])
                 
                 row1[0].write("**Model**")
                 row1[1].write("**Configuration**")
                 row1[2].write("**Audio**")
                 
                 row2[0].write(f"Meta MMS")
-                row2[1].write("[khof312 - femalehttps://huggingface.co/khof312/mms-tts-lin-female)") 
+                row2[1].write("[khof312 - female](https://huggingface.co/khof312/mms-tts-lin-female)") 
                 row2[2].audio(finetuned_mms1[0], sample_rate = finetuned_mms1[1]) 
+
+            
+                row3[0].write(f"African voices")
+                row3[1].write("[African Voices]()") 
+                row3[2].audio(finetuned_africanvoices[0], sample_rate = finetuned_africanvoices[1]) 
         
             st.divider()
                         

src/lookups.py

@@ -82,7 +82,8 @@ models = {
     'coqui': 'tts_models/lin/openbible/vits', # Sampling rate: 22050
     'espeakng': None,
     'toucan': 'Lingala (lin)',
-    'piper': None
+    'piper': None,
+    'africanvoices': 'cmu_lin_ope',
 }, 
 'mos':{
     'mms': 'facebook/mms-tts-mos',

src/synthesize.py

@@ -88,7 +88,31 @@ def synth_espeakng(text:str, model:str):
         return wav, sampling_rate
     else:
         return None
+        
+def synth_africanvoices(text:str, model:str):
+    '''
+    Use ESpeak-NG to synthesize text.
+
+    Inputs:
+        text: Text to synthesze
+        model: Model code 
+    Returns:
+        Streaming Wav and sampling rate.
+    '''
+    if model is not None:
+        
+        subprocess.run(['flite', f'-voice {model}.flitevox', f'"{text}"', " test.wav"]) 
+        #esng = espeakng.Speaker()
+        #esng.voice = model
+        #esng.say(text, export_path="test.wav")
 
+        sampling_rate, wav = wavfile.read('test.wav')
+        os.remove("test.wav")
+        
+        #wav = tts.tts(text=text)
+        return wav, sampling_rate
+    else:
+        return None
 
 def synth_toucan(text:str, model:str):
     '''