Update app.py

app.py

@@ -1,84 +1,24 @@
 import gradio as gr
-from transformers import AutoTokenizer, AutoModelForCausalLM
-import torch
+from transformers import pipeline
 
-# Load model
-model_name = "apple/FastVLM-0.5B"
-tokenizer = AutoTokenizer.from_pretrained(model_name)
-model = AutoModelForCausalLM.from_pretrained(model_name)
+# Load MMS-TTS pipeline
+tts = pipeline("text-to-speech", model="facebook/mms-tts")
 
-if tokenizer.pad_token is None:
-    tokenizer.pad_token = tokenizer.eos_token
-
-def generate_response(user_input):
-    # Fixed prompt template
-    fixed_prompt = """
-You are Hypr, an artificial intelligence assistant. 
-Your goal is to help users understand their surroundings, manage their lives better, and make decisions that improve life for future generations.
-
-When responding to a user's question:
-- Answer clearly, completely, and in full sentences.
-- Do not omit important words or information.
-- Keep your tone friendly, intelligent, and helpful.
-- If unsure, provide the most likely answer based on context.
-- Avoid unnecessary repetition or irrelevant details.
-
-User question: {user_input}
-Hypr response:
-
-"""
-    
-    # Combine fixed prompt + user input
-    full_prompt = f"{fixed_prompt}\n\n### Instruction:\n{user_input}\n\n### Response:\n"
-    
-    inputs = tokenizer.encode(full_prompt, return_tensors="pt")
-    
-    with torch.no_grad():
-        outputs = model.generate(
-            inputs,
-             max_length=400,           # ✅ REDUCED from 400 to 150
-            temperature=0.7,
-            do_sample=True,
-            pad_token_id=tokenizer.eos_token_id,
-            repetition_penalty=1.15,   # ✅ INCREASED to reduce repetition
-            top_p=0.9,
-            top_k=50,                 # ✅ ADDED top-k sampling
-            no_repeat_ngram_size=2,   # ✅ PREVENT repeating phrases
-            early_stopping=True       # ✅ STOP early when complete
-        )
-    
-    full_text = tokenizer.decode(outputs[0], skip_special_tokens=True)
-    
-    # Extract only the response part
-    if "### Response:" in full_text:
-        response = full_text.split("### Response:")[1].strip()
-        if "### Instruction:" in response:
-            response = response.split("### Instruction:")[0].strip()
-    else:
-        response = full_text
-    
-    return response
+# Function to convert text to speech
+def text_to_speech(text):
+    audio = tts(text)
+    # Save audio to file for Gradio output
+    with open("output.wav", "wb") as f:
+        f.write(audio["audio"])
+    return "output.wav"
 
 # Gradio interface
 iface = gr.Interface(
-    fn=generate_response,
-    inputs=gr.Textbox(
-        label="Ask Hypr AI", 
-        placeholder="Hello Hypr, can you help me with...",
-        lines=3
-    ),
-    outputs=gr.Textbox(
-        label="Hypr's Response", 
-        lines=6,
-        show_copy_button=True
-    ),
-    title="🤖 Hypr AI Assistant",
-    description="Meet Hypr - Your helpful AI assistant ready to help with any task!",
-    examples=[
-        ["Can you help me solve this math problem: 2 + 2 = ?"],
-        ["What is the capital of France?"],
-        ["How to make a cup of tea?"]
-    ]
+    fn=text_to_speech,
+    inputs=gr.Textbox(lines=2, placeholder="Type your text here..."),
+    outputs=gr.Audio(type="filepath"),
+    title="Multilingual TTS with MMS-TTS",
+    description="Type text in any supported language and get speech output."
 )
 
-iface.launch(share=False, debug=True) 
+iface.launch()