Update app.py

app.py

@@ -10,27 +10,25 @@ import os
 from huggingface_hub import snapshot_download
 
 # ... (Whisper and Vicuna setup remain the same)
-
 # --- VITS (TTS) Setup ---
-TTS_MODEL_NAME = "espnet/kan_bayashi_ljspeech_vits"
+TTS_MODEL_NAME = "espnet/speechlm_tts_v1"  # Updated Model Name
 tts_device = "cuda" if torch.cuda.is_available() else "cpu"
 
-# Download the ESPnet model files and get the download path
-model_dir = "vits_model"
-if not os.path.exists(model_dir):
-    os.makedirs(model_dir)
+model_dir = "speechlm_model" # Updated directory name
+
+if os.path.exists(model_dir):
+    shutil.rmtree(model_dir)
 
+os.makedirs(model_dir)
 download_path = snapshot_download(repo_id=TTS_MODEL_NAME, local_dir=model_dir, local_dir_use_symlinks=False)
-print(f"Downloaded ESPnet model to: {download_path}") # Print the path!
+print(f"Downloaded ESPnet model to: {download_path}")
 
-# Construct *absolute* paths to the config and model files.
-config_path = os.path.join(download_path, "exp/tts_train_vits_raw_phn_tacotron_g2p_en_no_space/config.yaml")
-model_path = os.path.join(download_path, "exp/tts_train_vits_raw_phn_tacotron_g2p_en_no_space/train.total_count.ave_10best.pth")
+# --- KEY CHANGE: Adjust paths for speechlm_tts_v1 ---
+config_path = os.path.join(download_path, "exp/speechlm_tts_v1/config.yaml")  # Correct path for speechlm_tts_v1
+model_path = os.path.join(download_path, "exp/speechlm_tts_v1/model.pth")  # Correct path for speechlm_tts_v1
 
-# Load the Text2Speech model using the downloaded files and absolute paths
 tts_model = Text2Speech(train_config=config_path, model_file=model_path, device=tts_device)
 
-
 # --- Vicuna (LLM) Setup ---
 VICUNA_MODEL_NAME = "lmsys/vicuna-7b-v1.5"
 vicuna_device = "cuda" if torch.cuda.is_available() else "cpu"