Add token authentication from space secrets and initial Piper synthesis code.

src/synthesize.py

@@ -29,7 +29,7 @@ def synth_mms(text:str, model:str):
     #                    raw_response=True)._content
 
     if model is not None:
-        pipe = pipeline("text-to-speech", model=model, device=-1) # Change device if it should use GPU
+        pipe = pipeline("text-to-speech", model=model, device=-1, token=os.environ['TOKEN']) # Change device if it should use GPU
         mms_tts = pipe(text)
         return mms_tts['audio'], mms_tts['sampling_rate']
     else:
@@ -100,7 +100,7 @@ def synth_toucan(text:str, model:str):
     Returns:
         Streaming Wav and sampling rate.
 
-    NOTE: This wrapper does not let you explore the full range of options possible with the API. The API should allow you to generate female voices, however, it does not seem to be working at the moment.
+    NOTES: (1)This wrapper does not let you explore the full range of options possible with the API. (2) The API should allow you to generate female voices, however, it does not seem to be working at the moment. (3) This uses a Huggingface Gradio Space to compute via the API.
     '''
     client = Client("Flux9665/MassivelyMultilingualTTS")
     result = client.predict(
@@ -115,4 +115,28 @@ def synth_toucan(text:str, model:str):
     		api_name="/predict"
     )
     sampling_rate, wav = wavfile.read(result[0])
+    return wav, sampling_rate
+
+def synth_piper(text:str, model:str):
+    '''
+    Use Toucan to synthesize text.
+
+    Inputs:
+        text: Text to synthesze
+        model: Model code 
+    Returns:
+        Streaming Wav and sampling rate.
+
+    NOTES: (1) This uses a Huggingface Gradio Space to compute via the API.
+    '''
+    client = Client("k2-fsa/text-to-speech")
+    result =  client.predict(
+    		language=model[0],
+    		repo_id=model[1],
+    		text=text,
+    		sid="0",
+    		speed=1,
+    		api_name="/process"
+    )
+    sampling_rate, wav = wavfile.read(result[0])
     return wav, sampling_rate