jj

app.py

@@ -9,15 +9,16 @@ from neon_tts_plugin_coqui import CoquiTTS
 # Whisper: Speech-to-text
 model = whisper.load_model("base")
 
-# The LLM : Bloom
+# LLM : Bloom as inference
 API_URL = "https://api-inference.huggingface.co/models/bigscience/bloom"
 HF_TOKEN = os.environ["HF_TOKEN"]
 headers = {"Authorization": f"Bearer {HF_TOKEN}"}
-#en, fr, esp, arb, hn, portu, Indonesian, Vietnamese, Chinese, tamil, telugu, bengali
+#Language covered in Bloom : en, fr, esp, arb, hn, portu, Indonesian, Vietnamese, Chinese, tamil, telugu, bengali
 
 # Text-to-Speech
 LANGUAGES = list(CoquiTTS.langs.keys())
 print(f"Languages for Coqui are: {LANGUAGES}")
+#Languages for Coqui are: ['en', 'es', 'fr', 'de', 'pl', 'uk', 'ro', 'hu', 'el', 'bg', 'nl', 'fi', 'sl', 'lv', 'ga']
 coquiTTS = CoquiTTS()
 
 # Whisper - speeech-to-text
@@ -46,12 +47,12 @@ def fun_engine(audio) :
   text1 = whisper_stt(audio)
   #text1 = model.transcribe(audio)["text"]
   text2 = lang_model_response(text1)
-  speech = tts(text, 'en')
+  speech = tts(text2, 'en')
   return text1, text2, speech
 
 # LLM - Bloom Response  
 def lang_model_response(prompt): 
-  print(f"*****Inside meme_generate - Prompt is :{prompt}")
+  print(f"*****Inside lang_model_response - Prompt is :{prompt}")
   if len(prompt) == 0:
     prompt = """Can you help me please?"""