New Features

• Custom Model Selector

• Session Notification
• Custom Model Path API (WIP)

Improvements

• Node Architecture Refactored - Monolithic node split into focused modules: tts_node, voice_cloning_node, advanced_params_node, with shared utilities in modules/ (custom_types, device_utils, dtype_utils, settings).

Bug Fixes

• AudioVAE device mismatch
• PyTorch 2.9+ compatibility

Migration Notes

• Need to update previous workflows