Convert to Gradio Space and optimize for ZeroGPU

Dockerfile

@@ -1,27 +0,0 @@
-FROM python:3.10-slim
-
-# Install system dependencies
-RUN apt-get update && apt-get install -y \
-    ffmpeg \
-    git \
-    && rm -rf /var/lib/apt/lists/*
-
-# Set working directory
-WORKDIR /app
-
-# Copy requirements and install
-COPY requirements.txt .
-RUN pip install --no-cache-dir -r requirements.txt
-
-# Copy application code
-COPY . .
-
-# Expose the Gradio port
-EXPOSE 7860
-
-# Set environment variables for Gradio
-ENV GRADIO_SERVER_NAME="0.0.0.0"
-ENV GRADIO_SERVER_PORT=7860
-
-# Run the application
-CMD ["python", "app.py"]

README.md

@@ -3,7 +3,7 @@ title: VoiceScript
 emoji: 🏆
 colorFrom: purple
 colorTo: gray
-sdk: docker
+sdk: gradio
 pinned: false
 ---
 

app.py

@@ -5,16 +5,18 @@ from faster_whisper import WhisperModel
 import torch
 import spaces
 
-# Initialize model
-# Note: On ZeroGPU, we initialize on CPU or wait for the GPU function
-device = "cuda" if torch.cuda.is_available() else "cpu"
-compute_type = "float16" if torch.cuda.is_available() else "int8"
+# Global cache for the model so we don't reload it if not necessary
+# But on ZeroGPU, it's safer to load it per request or rely on the container state.
+_cached_model = None
 
-print(f"Initial check - CUDA available: {torch.cuda.is_available()}")
-print(f"Loading Whisper Large V3...")
-
-# Global model variable
-model = WhisperModel("large-v3", device=device, compute_type=compute_type)
+def get_model():
+    global _cached_model
+    if _cached_model is None:
+        device = "cuda" if torch.cuda.is_available() else "cpu"
+        compute_type = "float16" if torch.cuda.is_available() else "int8"
+        print(f"Loading Whisper Large V3 on {device} ({compute_type})...")
+        _cached_model = WhisperModel("large-v3", device=device, compute_type=compute_type)
+    return _cached_model
 
 def format_timestamp(seconds):
     h = int(seconds // 3600)
@@ -39,7 +41,10 @@ def transcribe(audio_path, task="transcribe", language=None):
     if audio_path is None:
         return "Please upload an audio file.", None
 
-    print(f"Transcribing {audio_path} on {device}...")
+    # Get model inside the ZeroGPU context
+    model = get_model()
+    
+    print(f"Transcribing {audio_path}...")
     
     options = {
         "task": task,

packages.txt

@@ -0,0 +1 @@
+ffmpeg