v0.15.1
📦 comfyuiView on GitHub →
✨ 3 features🐛 7 fixes🔧 2 symbols
Summary
This release introduces support for the FlowRVS segmentation model and per-guide attention control, alongside numerous bug fixes related to prompt parsing, LTXAV memory estimation, and GLSL node texture acceleration. Dynamic VRAM is now disabled when using the torch compiler.
Migration Steps
- If encountering issues with dynamic VRAM, consider that it is now disabled when using the torch compiler.
✨ New Features
- Support WanVideo based segmentation model FlowRVS.
- Implement per-guide attention strength control in self-attention.
- Add NanoBanana2 node to api-nodes.
🐛 Bug Fixes
- Fix KeyError when prompt entries lack class_type key.
- Fix Aimdo fallback on probe to not use zero-copy sft.
- Patch frontend to 1.39.19 (fix progress text serialization in API nodes).
- Fix LTXAV text enc min length.
- Fix ltxav te mem estimation.
- Fix GLSL node with pyopengl accelerate numpy textures.
- Move essentials_category to correct replacement nodes.