Fix: Skip faster-whisper for GGUF files

faster-whisper doesn't support GGUF format (it's llama.cpp format).
Now detects GGUF files by extension and goes directly to whispercpp.

coderai

@@ -2383,7 +2383,15 @@ async def create_transcription(
         tmp_path = tmp.name
     
     try:
-        # Try faster-whisper first (requires PyTorch), fall back to whispercpp if it fails
+        # Check if model is a GGUF file - faster-whisper doesn't support GGUF format
+        is_gguf_model = model_to_use.endswith('.gguf') or 'gguf' in model_to_use.lower()
+        
+        if is_gguf_model:
+            # Skip faster-whisper for GGUF files - go directly to whispercpp
+            print("Detected GGUF model - using whispercpp backend")
+            faster_whisper_failed = True
+        else:
+            # Try faster-whisper first
             faster_whisper_failed = False
             try:
                 from faster_whisper import WhisperModel
@@ -2483,7 +2491,7 @@ async def create_transcription(
                 # faster-whisper not available, will try whispercpp below
                 faster_whisper_failed = True
             except Exception as e:
-            # faster-whisper failed for some other reason (e.g., GGUF file not supported)
+                # faster-whisper failed for some other reason
                 print(f"Warning: faster-whisper failed to load model: {e}")
                 faster_whisper_failed = True
         
@@ -3755,7 +3763,18 @@ def main():
         should_preload = load_mode in ("loadall", "loadswap") or (model_name is None and args.audio_model)
         if should_preload:
             print(f"Pre-loading audio model...")
-            # First try faster-whisper, then fall back to whispercpp if it fails
+            
+            # Check if model is a GGUF file - faster-whisper doesn't support GGUF format
+            model_to_use = args.audio_model
+            is_gguf_model = model_to_use.endswith('.gguf') or 'gguf' in model_to_use.lower()
+            
+            if is_gguf_model:
+                # Skip faster-whisper for GGUF files - it doesn't support them
+                # Go directly to whispercpp
+                print("Detected GGUF model - using whispercpp backend")
+                faster_whisper_failed = True
+            else:
+                # Try faster-whisper first
                 faster_whisper_failed = False
             try:
                 # Try faster-whisper first (requires torch)
@@ -3802,6 +3821,8 @@ def main():
             
             # If faster-whisper failed (not installed or couldn't load), try whispercpp
             if faster_whisper_failed:
+                # Check if model is a GGUF file - whispercpp can handle those
+                model_is_gguf = model_to_use.endswith('.gguf') or (model_path and model_path.endswith('.gguf'))
                 try:
                     import whispercpp