v6.14.2
umputun/ralphexv6.14.2Sep 27, 2026by lstein
Upgrading?
AI Summary
This release focuses on bug fixes and minor enhancements including custom font support for Canvas and image style transfer capabilities to krea-2. It addresses security vulnerabilities, improves model handling, and refines various UI and performance aspects.
Key Highlights
- Added custom font import support for Canvas text tool
- Image style transfer to krea-2 via reference image
- Fixed password-change token handling during routine refreshes
- Security fix for path traversal in object-serializer names
- Improved VRAM memory management for quantized models
New Features
- Custom font import for Canvas text tool
- Image style transfer to krea-2 using reference images
- Improved VRAM reservation for VAEs
- Enhanced internal memory management at shutdown
- Optimized VRAM management for Qwen2.5-VL encoder
Full Release Notes
This release fixes a number of bugs in version 6.14.1 and adds a couple of useful new features. ## New Features in this release * You can now import custom fonts for use with the Canvas text tool (@DustyShoe) * You can now transfer an image's style to krea-2 by providing a reference image (@Pfannkuchensack) ## Bug fixes * Keep a password-change token when a routine refresh lands first by @MohammadHijjawi97 in https://github.com/invoke-ai/InvokeAI/pull/9599 * Reject path traversal in object-serializer names and model keys by @lstein in https://github.com/invoke-ai/InvokeAI/pull/9595 * Clarify proxy destination policy without blocking local proxies by @wunianze666-netizen in https://github.com/invoke-ai/InvokeAI/pull/9554 * Accept epoch-changing token refreshes by @wunianze666-netizen in https://github.com/invoke-ai/InvokeAI/pull/9560 * Remove FP4 Mistral starter, fix model rename validation, limit Z-Image to Qwen3 4B encoders by @Pfannkuchensack in https://github.com/invoke-ai/InvokeAI/pull/9587 * Stop the SDNQ guards from aborting folder-encoder identification by @Pfannkuchensack in https://github.com/invoke-ai/InvokeAI/pull/9561 * img2img posterization - noise scaling and sigma schedule by @Pfannkuchensack in https://github.com/invoke-ai/InvokeAI/pull/9556 * Make I hotkey work in Canvas, Gallery, and Viewer by @DustyShoe in https://github.com/invoke-ai/InvokeAI/pull/9569 * Make room in VRAM for the quantized Qwen2.5-VL encoder and release it afterwards (#9147) by @lstein in https://github.com/invoke-ai/InvokeAI/pull/9572 * Buffer upload metadata in a bounded bytearray, not a list[bytes] by @lstein in https://github.com/invoke-ai/InvokeAI/pull/9583 * Ignore unexpected checkpoint keys, report them at debug only by @lstein in https://github.com/invoke-ai/InvokeAI/pull/9581 * Pin the scope of the socket's server-disconnect retry bound by @lstein in https://github.com/invoke-ai/InvokeAI/pull/9582 * Honor ComfyUI's comfy.gguf.orig_shape metadata by @Pfannkuchensack in https://github.com/invoke-ai/InvokeAI/pull/9564 ## New Contributors * @MohammadHijjawi97 made their first contribution in https://github.com/invoke-ai/InvokeAI/pull/9599 **Full Changelog**: https://github.com/invoke-ai/InvokeAI/compare/v6.14.1...v6.14.2
Related Repositories
Get up and running with Kimi, GLM, MiniMax, DeepSeek, gpt-oss, Qwen, Gemma and other models.
A fast reverse proxy to help you expose a local server behind a NAT or firewall to the internet.
RAGFlow is a leading open-source Retrieval-Augmented Generation (RAG) engine that fuses cutting-edge RAG with Agent capabilities to create a superior context layer for LLMs
The world’s fastest framework for building websites.
:cherry_blossom: A command-line fuzzy finder