Update app.py

app.py

@@ -5,7 +5,7 @@ import os
 
 print("🚀 Starting Kurmanji ASR application...")
 
-# Define MODEL_ID. Assuming it's in the root of the Space, use the Space ID.
+# Define MODEL_ID: Set to '.' because model files are in the same Space repository.
 MODEL_ID = "." 
 
 # --- MODEL LOADING FUNCTION ---
@@ -15,16 +15,18 @@ def load_asr_model():
     try:
         print("📥 Loading ASR pipeline...")
         
+        # Use the pipeline's auto-loading feature
         asr = pipeline(
             "automatic-speech-recognition",
             model=MODEL_ID,
-            chunk_length_s=5,          # Reduced for stability
-            stride_length_s=(1, 1),    # Reduced for stability
-            device="cpu",              # Corrected device specification
-            low_cpu_mem_usage=True,    # Added for memory efficiency
-            # Ensure this line ends with a comma (,) and the parenthesis is closed on the next line
-            generate_kwargs={"language": "ku", "task": "transcribe"}  
-        ) # <-- ENSURE THIS CLOSING PARENTHESIS IS HERE
+            # Optimized for limited resources/speed—you may need to adjust these values
+            chunk_length_s=5,          # Smaller chunks require less memory
+            stride_length_s=(1, 1),    # Reduced stride for less overlap
+            device="cpu",              # Explicitly use CPU (important for free tier)
+            low_cpu_mem_usage=True,    # Saves RAM during model initialization
+            # VITAL FOR QUALITY: Force the model to use the Kurdish language (ku)
+            generate_kwargs={"language": "ku", "task": "transcribe"} 
+        )
         print("✅ ASR pipeline created successfully!")
     except Exception as e:
         print(f"❌ Error loading ASR model: {e}")
@@ -36,7 +38,6 @@ load_asr_model() # Load the model at start
 # --- TRANSCRIPTION FUNCTION ---
 def transcribe(audio_file):
     print("=== ASR Function Called ===")
-    print(f"Audio file: {audio_file}")
     
     try:
         if audio_file is None:
@@ -45,10 +46,13 @@ def transcribe(audio_file):
         if asr is None:
             return "Model nehatiye barkirin. / ASR model not loaded properly."
         
-        print("🎵 Processing audio file...")
+        print(f"🎵 Processing audio file: {audio_file}")
         
+        # Transcribe the audio
         result = asr(audio_file)
-        transcription = result["text"].strip() # Clean the output
+        
+        # Post-processing: clean output and remove leading/trailing spaces
+        transcription = result["text"].strip() 
 
         print(f"✅ Transcription completed: {transcription}")
         return transcription
@@ -86,5 +90,4 @@ demo = gr.Interface(
 # --- LAUNCH ---
 print("🚀 Launching Gradio app...")
 if __name__ == "__main__":
-    # FIX APPLIED HERE: Removed 'ssr_mode=False'
     demo.launch()