December 2025 Summaries

ComfyUI is transitioning its repository from the @comfyanonymous account to the @Comfy-Org organization to better support its growth and streamline internal workflows. This change, effective by January 6, aims to enhance collaboration and security by allowing for improved management of permissions and access to advanced security tools. Users are assured that GitHub will automatically redirect all existing links to the new repository location, although updating local git remotes is recommended. This organizational shift is designed to facilitate easier integration of automation and AI tools, ultimately making it more straightforward for contributors to participate in the project. The move underscores ComfyUI's commitment to open-source development and aims to foster greater community involvement, faster iteration, and a healthier pull request and review process, with plans to establish a long-term open governance structure as the project scales.

ComfyUI now offers native support for Tripo v3.0, integrating the latest advancements in 3D asset creation directly into its workflow. Tripo v3.0 enhances geometry accuracy and texture quality with its redesigned pipeline, providing sharper edges, coherent structures, and higher texture resolutions, alongside improved PBR behavior for more natural color and light realism. Users can choose between Standard and Detailed (Ultra) modes, allowing a balance between real-time efficiency and high-fidelity output, with Ultra mode enabling up to 2 million faces for complex assets. This integration allows users to generate production-ready 3D assets without leaving ComfyUI, simplifying the process of creating high-quality 3D outputs.

Qwen Image Edit 2511 and Qwen Image Layered are two advanced image-editing models integrated into ComfyUI, offering high-fidelity, instruction-driven editing and structured, layer-aware composition for enhanced creative control and compositing. Qwen Image Edit 2511 is an improved version of its predecessor, featuring better consistency and capabilities for instruction-driven image editing, character consistency, and localized control, with built-in support for community-created LoRAs. Meanwhile, Qwen Image Layered facilitates the decomposition of images into multiple RGBA layers, allowing for independent manipulation of components, which is ideal for designers and VFX artists working in modular image pipelines. Both models can be accessed through the Comfy Cloud platform, enabling users to update to the latest ComfyUI version and download necessary workflows for seamless integration into their projects.

Recent updates to ComfyUI include enhancements to the Template Library and new support for advanced image and video models like GPT-Image-1.5, Kling 2.6, and Wan 2.6. The updated Template Library now features workflows designed for practical applications and creative projects, which are open-sourced and can be integrated directly into local ComfyUI setups. GPT-Image-1.5 improves image editing by providing faster rendering, better detail preservation, and consistent lighting, while Kling 2.6 introduces visuals with integrated audio, creating a seamless blend of sound and imagery. Wan 2.6 expands capabilities with character-driven, multi-shot storytelling videos and synchronized audio, offering studio-quality visual narratives. These developments demonstrate ComfyUI's commitment to innovation and responsiveness to community feedback, aiming to enhance user creativity and project execution.

The latest update to ComfyUI introduces the ComfyUI-Manager, a newly integrated extension management system designed to streamline workflows and improve node management efficiency. This update features a revamped user interface aimed at quicker discovery, safer installations, and easier management of extensions, including capabilities like pre-installation previews of node information, batch installation of missing nodes, and early detection of dependency conflicts with visual alerts. Enhanced security measures have been added, such as scanning nodes for potential threats and providing security warnings, while a better search function and full localization support improve user experience. Desktop users have the new UI enabled by default, whereas other users can activate it by installing the ComfyUI-Manager pip version and adjusting launch commands. The update also maintains compatibility with the legacy UI, allowing users to switch back if desired. The data migration warning informs users about the relocation of Manager data directories to enhance protection against unauthorized access, and users are encouraged to provide feedback to help further refine the extension management experience.

In a bid to improve the ComfyUI platform in 2026, the team has launched a user survey to gather feedback from the community, reflecting on the developments of the past year, which included the introduction of Comfy Cloud (Beta), expanded support for open-source models and Partner Nodes, and the release of Nodes 2.0 and Subgraph. The survey aims to gain insights into user experiences, understand what aspects of the platform work well, identify areas needing improvement, and help shape the future roadmap by prioritizing new features. Participants have until January 1st, 2026, to complete the survey, which takes around five minutes, with 30 randomly selected respondents receiving a month of the Comfy Cloud Standard Plan and exclusive merchandise as a token of appreciation. The feedback will be instrumental in guiding the strategic direction of ComfyUI, ensuring it continues to evolve in alignment with user needs and preferences.

Comfy Cloud has announced a significant update that introduces new features and a revised pricing model to enhance flexibility and scalability for creators. Users can now import models from platforms like Civitai directly into Comfy Cloud, with HuggingFace support forthcoming, and utilize upgraded Blackwell RTX 6000 Pro GPUs, which offer improved performance for complex workflows. Additionally, workflow runtimes have been extended to up to one hour for Pro plan users, and a new template library with creative, shareable workflows has been introduced. The service has transitioned to a unified credit-based system, which applies to both Cloud and Partner Node usage, and three new subscription plans—Standard, Creator, and Pro—are now available, allowing users to scale their usage based on their needs. Early subscribers to the service are rewarded with a Founder’s Edition upgrade, which includes bonus credits, to ensure continued access to enhanced features. These changes are designed to support the growing demands of advanced workflows while maintaining performance and stability.

Ubisoft La Forge has open-sourced its CHORD model and ComfyUI nodes for generating PBR materials, offering an end-to-end solution for AAA game production pipelines. The release, under a Research-Only license, aims to integrate AI-driven workflows into large-scale video game development, addressing production bottlenecks in creating high-quality, reusable materials that meet svBRDF standards. The CHORD model is part of a three-stage pipeline consisting of texture image generation, image-to-material estimation, and material upscaling, designed to streamline the creation of PBR texture sets for real-time game production. Ubisoft's choice of ComfyUI as the workflow platform is due to its efficiency in synthesizing textures, estimating materials, and upscaling, while providing fine-grain control through features like ControlNets and inpainting. This open-source initiative not only facilitates rapid iteration for artists but also encourages the integration of AI tools into existing production workflows, highlighting a commitment to open foundations in industrial graphics AI development.

Kling O1, a new feature available in ComfyUI as a partner node, is based on the Multi-modal Visual Language (MVL) concept and marks a new era in video editing by enhancing video generation models. It offers a comprehensive suite of capabilities, allowing users to input videos and use text prompts for edits, such as modifying camera angles, adding or removing objects, and changing backgrounds. The tool supports both text-to-video and image-to-video functionalities, enabling complex edits with multiple character or object references while maintaining consistency. Users can input up to one video and four images together, with specific requirements for resolution and file size, to leverage features like video inpainting and green screen keying. To begin using Kling O1, users can either try it on the Comfy Cloud or update their local ComfyUI to the latest version, accessing the Kling Omni Edit Model nodes or the Kling O1 workflow through the Template Library.

ComfyUI has introduced a significant update to its interface with Nodes 2.0, transitioning from LiteGraph.js Canvas rendering to a Vue-based architecture to enhance speed and interaction capabilities. This shift addresses limitations of the previous system, allowing for more powerful feature development, expandable nodes, and improved interaction patterns, while maintaining the option to revert to the legacy Canvas if needed. The update prioritizes seamless migration for custom nodes, ensuring that the community's contributions remain integral. Active work is underway to address user feedback, including restoring features like Stop/Cancel and optimizing performance. ComfyUI emphasizes a fast-paced iteration philosophy, encouraging user feedback to refine developments and ensuring a collaborative evolution of the platform.