add hf space config

README.md

@@ -1,3 +1,28 @@
+---
+title: BARK Text to Audio with Batch Inference
+emoji: 🪄
+colorFrom: purple
+colorTo: pink
+sdk: gradio
+python_version: "3.12"
+sdk_version: "5.23.3"
+suggested_hardware: cpu-upgrade
+suggested_storage: small
+app_file: app.py
+short_description: Generate natural sounding speech audio from text
+pinned: true
+startup_duration_timeout: 45m
+# models:
+#     - mistralai/Mistral-7B-Instruct-v0.2
+#     - stabilityai/stable-diffusion-xl-base-1.0
+tags:
+    - text-to-audio
+    - gradio
+    - bark
+preload_from_hub:
+    - suno/bark coarse.pt coarse_2.pt fine.pt fine_2.pt text.pt text_2.pt
+---
+
 # Generate Audio from text and clone voice with BARK
 
 You can generate audio from text with natural sounding voice and clone any voice (not perfect).

core/memory/common.py

@@ -57,7 +57,7 @@ class EnvVars:
         "GLOBAL_ENABLE_MPS": True,  # Enable PyTorch's Metal Performance Shaders on Apple Silicon
         "AUDIO_SAMPLE_RATE": 24000,  # Default sample rate for audio processing (in Hz)
         "SUNO_USE_SMALL_MODELS": True,  # Use smaller Bark models if True
-        "CACHE_DIR": "./models",
+        "CACHE_DIR": "~/.cache/huggingface/hub",
         "LOG_LEVEL": LogLevel.INFO,  # Default logging level
         "USE_GPU": True,  # Whether to prioritize GPU/MPS over CPU
     }