v0.27.0
📦 comfyuiView on GitHub →
✨ 14 features🐛 3 fixes🔧 2 symbols
Summary
This release introduces major support for int8 models, bringing performance improvements and bug fixes related to int8 operations, alongside several new partner node integrations and feature additions like asset hashing control.
Migration Steps
- If using asset hashing, enable it explicitly using the --enable-asset-hashing flag, as it is now opt-in.
✨ New Features
- Added support for int8 convrot models.
- Added support for HappyHorse 1.1 model (Partner Nodes - Alibaba).
- Added 1080p resolution to Grok Image node (Partner Nodes - Grok).
- Added 4K resolution support for SeeDance 2.0 (Partner Nodes - ByteDance).
- Added support for Seed node (CORE-295).
- Added advanced krea 2 model merging node.
- Added Bounding box canvas and Ideogram JSON prompt support (CORE-292).
- Added support for SeeDance-2.0-Mini video model (Partner Nodes - ByteDance).
- Updated GLSL node to use ANGLE library (CORE-162).
- Enabled int8 support on turing GPUs.
- Added ConditioningMultiply node to nodes.py.
- Added Nano Banana 2 Lite model support (Partner Nodes - Google).
- Added Gemini Video Omni node support (Partner Nodes - Google).
- Made asset hashing opt-in via --enable-asset-hashing (off by default).
🐛 Bug Fixes
- Fixed int8 loras causing lower quality requant with wrong settings.
- Fixed int8 mm being skipped on offloaded lora weights.
- Fixed memory leak related to int8.