Update app.py

app.py

@@ -4,13 +4,13 @@ import torch
 import tempfile
 import asyncio
 import edge_tts
-import spaces
+from pydub import AudioSegment
 from threading import Thread
 from collections.abc import Iterator
 from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
 
 DESCRIPTION = """
-# QwQ Tiny with Edge TTS
+# QwQ Tiny with Edge TTS (MP3 Output)
 """
 
 MAX_MAX_NEW_TOKENS = 2048
@@ -29,14 +29,23 @@ model = AutoModelForCausalLM.from_pretrained(
 model.eval()
 
 async def text_to_speech(text: str) -> str:
-    """Converts text to speech using Edge TTS and returns the generated audio file path."""
+    """Converts text to speech using Edge TTS, converts WAV to MP3, and returns the MP3 file path."""
+    with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp_wav:
+        wav_path = tmp_wav.name
+    
     communicate = edge_tts.Communicate(text)
-    with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp_file:
-        tmp_path = tmp_file.name
-        await communicate.save(tmp_path)
-    return tmp_path
+    await communicate.save(wav_path)
+
+    # Convert WAV to MP3
+    with tempfile.NamedTemporaryFile(delete=False, suffix=".mp3") as tmp_mp3:
+        mp3_path = tmp_mp3.name
+    
+    audio = AudioSegment.from_wav(wav_path)
+    audio.export(mp3_path, format="mp3")
+
+    os.remove(wav_path)  # Delete the original WAV file
+    return mp3_path  # Return the MP3 file path
 
-@spaces.GPU
 def generate(
     message: str,
     chat_history: list[dict],
@@ -47,12 +56,12 @@ def generate(
     repetition_penalty: float = 1.2,
 ) -> Iterator[str] | str:
     
-    is_tts = message.strip().startswith("@tts")
+    is_tts = message.strip().startswith("edgetts@tts")
     is_text_only = message.strip().startswith("@text")
 
     # Remove special tags
     if is_tts:
-        message = message.replace("@tts", "").strip()
+        message = message.replace("edgetts@tts", "").strip()
     elif is_text_only:
         message = message.replace("@text", "").strip()
 
@@ -91,7 +100,7 @@ def generate(
         loop = asyncio.new_event_loop()
         asyncio.set_event_loop(loop)
         audio_path = loop.run_until_complete(text_to_speech(final_output))
-        return audio_path  # Returning audio file path
+        return audio_path  # Returning MP3 file path
 
     return final_output  # Returning text output
 
@@ -107,8 +116,8 @@ demo = gr.ChatInterface(
     stop_btn=None,
     examples=[
         ["A train travels 60 kilometers per hour. If it travels for 5 hours, how far will it travel in total?"],
-        ["@text What causes rainbows to form?"],
-        ["@tts Explain Newton's third law of motion."],
+        ["@text What is AI?"],
+        ["edgetts@tts Explain Newton's third law of motion."],
         ["@text Rewrite the following sentence in passive voice: 'The dog chased the cat.'"],
     ],
     cache_examples=False,