Update app.py

app.py

@@ -2,29 +2,8 @@ import gradio as gr
 import edge_tts
 import asyncio
 import tempfile
-
-import torch
-import soundfile as sf
-
-from melo import TTS
-
-
-# -----------------------------
-# Load MeloTTS
-# -----------------------------
-
-print("Loading MeloTTS...")
-
-device = "cuda" if torch.cuda.is_available() else "cpu"
-
-melo_tts = TTS(
-    language="EN",
-    device=device
-)
-
-speaker_ids = list(melo_tts.hps.data.spk2id.keys())
-
-print("MeloTTS loaded on:", device)
+import subprocess
+import os
 
 
 # -----------------------------
@@ -37,6 +16,7 @@ async def get_voices():
 
     return {
         f"{v['ShortName']} - {v['Locale']} ({v['Gender']})": v["ShortName"]
+        for v["ShortName"]
         for v in voices
     }
 
@@ -77,18 +57,14 @@ async def edge_tts_engine(text, voice, rate, pitch):
 
 
 # -----------------------------
-# MeloTTS
+# Mimic3 TTS
 # -----------------------------
 
-def melo_tts_engine(text, rate):
+def mimic3_tts_engine(text):
 
     if not text.strip():
         return None, "Please enter text."
 
-    speed = 1.0 + rate / 100
-
-    speaker = speaker_ids[0]
-
     with tempfile.NamedTemporaryFile(
         delete=False,
         suffix=".wav"
@@ -96,12 +72,15 @@ def melo_tts_engine(text, rate):
 
         path = tmp.name
 
-        melo_tts.tts_to_file(
-            text=text,
-            speaker_id=speaker,
-            file_path=path,
-            speed=speed,
-        )
+    # Run Mimic3 CLI
+    cmd = [
+        "mimic3",
+        "--voice", "en_US/vctk_low",
+        text,
+        "--output-file", path,
+    ]
+
+    subprocess.run(cmd, check=True)
 
     return path, None
 
@@ -118,12 +97,9 @@ async def unified_tts(
     engine,
 ):
 
-    if engine == "Melo (Open Source)":
+    if engine == "Mimic3 (Open Source)":
 
-        return melo_tts_engine(
-            text,
-            rate,
-        )
+        return mimic3_tts_engine(text)
 
     else:
 
@@ -148,7 +124,7 @@ async def create_demo():
         gr.Markdown("""
         # 🎙️ AI Text-to-Speech Lab (HF Free)
 
-        **MeloTTS (Open Source) vs Edge (Cloud)**
+        **Mimic3 (Open Source) vs Edge (Cloud)**
         """)
 
         text_input = gr.Textbox(
@@ -157,8 +133,8 @@ async def create_demo():
         )
 
         engine_radio = gr.Radio(
-            ["Melo (Open Source)", "Edge (Cloud)"],
-            value="Melo (Open Source)",
+            ["Mimic3 (Open Source)", "Edge (Cloud)"],
+            value="Mimic3 (Open Source)",
             label="Engine",
         )
 
@@ -167,12 +143,6 @@ async def create_demo():
             label="Edge Voice",
         )
 
-        rate_slider = gr.Slider(
-            -40, 40, 0,
-            step=1,
-            label="Speed",
-        )
-
         generate_btn = gr.Button("Generate", variant="primary")
 
         audio_output = gr.Audio()
@@ -184,7 +154,7 @@ async def create_demo():
             inputs=[
                 text_input,
                 voice_dropdown,
-                rate_slider,
+                None,
                 None,
                 engine_radio,
             ],