Update app.py

app.py

@@ -1,48 +1,28 @@
-import spaces
-@spaces.GPU(duration=15)
-def gpu():
-    print("[GPU] | GPU maintained.")
+# app.py
+import spaces, os, gradio as gr, torch
+from huggingface_hub import snapshot_download
+from indextts.infer import IndexTTS
 
-import os
-import sys
-import subprocess
-import multiprocessing
-import gradio as gr
+model_dir = snapshot_download("IndexTeam/IndexTTS-1.5", local_dir="checkpoints", local_dir_use_symlinks=False)
+cfg_path  = os.path.join(model_dir, "config.yaml")
+tts = None
 
-bitnet_dir = os.path.join(os.getcwd(), "bitnet")
-if not os.path.isdir(bitnet_dir):
-    subprocess.run(["git", "clone", "--depth", "1", "--recursive", "https://github.com/microsoft/BitNet.git", "bitnet"], check=True)
-build_dir = os.path.join(bitnet_dir, "build")
-if not os.path.isdir(build_dir):
-    os.makedirs(build_dir, exist_ok=True)
-    subprocess.run(["cmake", ".."], cwd=build_dir, check=True)
-    subprocess.run(["cmake", "--build", ".", "--config", "Release", "--parallel", str(multiprocessing.cpu_count())], cwd=build_dir, check=True)
+@spaces.GPU
+def synth(audio_path, text):
+	global tts
+	if tts is None:
+		tts = IndexTTS(model_dir=model_dir, cfg_path=cfg_path)
+		if torch.cuda.is_available(): tts.to("cuda")
+	out_path = "output.wav"
+	tts.infer(audio_path, text, out_path)
+	return out_path
 
-script_path = os.path.join(bitnet_dir, "run_inference.py")
-model_path = os.environ.get("MODEL_PATH", "models/BitNet-b1.58-2B-4T/ggml-model-i2_s.gguf")
-
-def generate(prompt, max_tokens=128, temperature=0.7):
-    cmd = [
-        sys.executable,
-        script_path,
-        "-m", model_path,
-        "-p", prompt,
-        "-n", str(max_tokens),
-        "-temp", str(temperature)
-    ]
-    proc = subprocess.run(cmd, cwd=bitnet_dir, capture_output=True, text=True)
-    return proc.stdout.strip() if proc.returncode == 0 else proc.stderr.strip()
-
-iface = gr.Interface(
-    fn=generate,
-    inputs=[
-        gr.Textbox(lines=2, placeholder="enter your prompt here", label="prompt"),
-        gr.Slider(1, 512, value=128, step=1, label="max tokens"),
-        gr.Slider(0.0, 1.0, value=0.7, step=0.01, label="temperature")
-    ],
-    outputs=gr.Textbox(label="completion"),
-    title="bitnet.cpp completion demo"
-)
+with gr.Blocks() as demo:
+	gr.Markdown("# index-tts 1.5 (english)")
+	text_in  = gr.Textbox(label="text prompt")
+	audio_in = gr.Audio(label="reference voice", type="filepath")
+	audio_out= gr.Audio(label="generated speech", type="filepath")
+	gr.Button("generate").click(synth, [audio_in, text_in], audio_out)
 
 if __name__ == "__main__":
-    iface.launch()
+	demo.launch()