2 Commits

Include renames

Author	SHA1	Message	Date
diodiogod	5b5f8d603c	Version 4.6.13 ... Fix MelBandRoFormer architecture compatibility issues Technical details: - Implement proper MelBandRoFormer model loading with correct config - Add Kijai's safetensors model support and auto-download - Fix tensor size mismatch in instrumentals calculation - Add model-specific output inversion handling - Prevent incompatible RVC fallback for MelBand models	2025-08-30 10:53:22 -03:00
diodiogod	d082c2850b	Enhance vocal removal with comprehensive guide and improved UX ... - Rename node to "🤐 Noise or Vocal Removal" for clearer functionality - Change output names to "extracted voice/noise/echo" and "remaining" for better clarity - Add comprehensive vocal removal guide in docs/VOCAL_REMOVAL_GUIDE.md - Update README with guide reference and feature description - Create concise tooltip referencing complete documentation - Fix automatic output swapping for inverted models (RoFormer, HP-Karaoke, DeEcho) - Enable GPU acceleration with onnxruntime-gpu by default - Add model_data.json to MDX downloads for proper configuration - Improve model filtering to exclude non-model files - Add audio-separator dependency for better model compatibility - Fix NumPy 2.x compatibility and tensor handling issues - Add PyTorch weights_only=False for compatibility	2025-08-08 11:10:42 -03:00