January 2026 Summaries

Grok Imagine, a generative media model family from xAI, is now available in ComfyUI, offering capabilities in text-to-image, image editing, text-to-video, image-to-video, and video editing, all currently in beta. This integration allows creators to use Grok Imagine's models as partner nodes within ComfyUI, providing a platform to manage instability and enhance creative control by combining Grok outputs with deterministic tools. Grok Imagine excels in fast, high-quality image generation, particularly in cinematic and moody styles with strong facial consistency and expressive lighting, as well as in creating anime and cyberpunk aesthetics. It also performs well in video generation, delivering expressive cyberpunk visuals and effective scene reconstruction from new perspectives. The ComfyUI team appreciates Grok Imagine's natural affinity for moody visual language, characterized by subdued color palettes and dramatic contrast, which enriches the creative possibilities for users.

ComfyUI has introduced day-0 native support for Z-Image, a foundational image-generation model designed to enhance creative control and allow for community fine-tuning. The non-distilled version of Z-Image offers richer visual details and a higher artistic ceiling, though it requires more steps for optimal quality compared to its distilled counterpart, Z-Image-Turbo, which prioritizes speed. This model supports diverse aesthetics, effective negative prompts, and enhanced generation diversity, making it an ideal base for specialized development. Users can access Z-Image workflows through ComfyUI's template library, with examples showcasing its ability to produce highly detailed and artistically varied images, such as photorealistic portraits, retro film photographs, and sophisticated interior designs. The model's capabilities are highlighted in various creative settings, demonstrating its potential for producing both high-quality and diverse visual content.

Vidu Q2, the latest video generation model from Vidu, is now integrated into ComfyUI, offering improved character consistency, enhanced speed, and support for workflows involving up to seven reference subjects. It emphasizes controllability and stability, enabling users to produce more consistent outputs across iterations. Key features include stronger multimodal understanding for better handling of complex instructions and emotions, higher-fidelity dynamic rendering for smoother motions, and refined expressions and micro-movements for greater character expressiveness. The model also provides advanced camera language control for coherent transitions and improved multi-character and scene coordination. Users can access various workflows, such as converting text or images to video, through Comfy Cloud by updating to the latest ComfyUI version and utilizing the available templates.

Seedance 1.5 Pro, an advanced audio-visual co-generation model by ByteDance, is now integrated into ComfyUI, offering users the ability to generate videos with synchronized dialogue, music, and sound effects across multiple languages, complete with automatic lip-sync. The model allows for frame-to-frame control, enabling users to define opening and closing frames for guiding the visual style and ensuring smooth transitions, while also delivering natural character expressions and emotional delivery to achieve cinematic quality. Users can explore various workflows such as text-to-video, image-to-video, and first-last-frame to video by updating their ComfyUI to the latest version and accessing the templates in the library, facilitating the creation of professional audio-visual content.

FIBO Edit, a JSON-native image editing model developed by Bria AI, is now integrated into ComfyUI, offering precise and targeted image modifications using structured JSON prompts. The model allows for specific changes in attributes such as lighting, color, and texture without affecting the rest of the image, making it ideal for commercial use due to its reliance on licensed data. FIBO Edit ensures reproducible and auditable edits, providing strong legal clarity for enterprises and agencies using ComfyUI in production. The tool's enterprise-grade foundation supports intelligent edit suggestions and robust prompt adherence, catering to professional teams that require high control and predictability. Users can explore various creative options like changing the time of day, swapping materials, or adjusting text, with the recent release of Bria Image Edit on Comfy Cloud and open weights on platforms like HuggingFace.

Meshy 6 is now integrated into ComfyUI, enabling users to generate high-quality 3D models with improved geometry and mesh quality. This update focuses on enhancing structural accuracy and providing users with the ability to create smoother, anatomically correct geometry for characters, as well as sharper edges and cleaner structures for mechanical models. It also supports low-poly optimization for efficient wireframes suitable for real-time performance. Users can access these features by updating to the latest version of ComfyUI and utilizing the Meshy 6 workflow templates available in the 3D category of the Templates Library. The integration facilitates workflows such as text-to-model, image-to-model, and multiview-to-model, allowing for diverse and production-ready 3D asset creation.

Alibaba's WAN 2.6 Reference-to-Video feature has been integrated into ComfyUI, enabling users to generate new cinematic videos by learning motion, camera behavior, and visual style from reference clips. This update supports creating short videos by using one or two reference videos alongside a text prompt to replicate their motion, framing, and style, producing outputs with resolutions up to 1080p in either portrait or landscape orientation. Users can control actions and scenes through prompts, and they can access this feature by updating ComfyUI, navigating to the Workflow Library, selecting Video, and choosing WAN 2.6 Reference to Video to upload clips and initiate the workflow.

FLUX.2 [klein] 4B & 9B models are the latest in the Flux family, providing high-speed image editing and generation with a focus on interactive workflows and immediate previews. These models come in two types—Base (Undistilled) and Distilled—offering flexibility and control for different applications. The Base models are optimized for fine-tuning and customization, while the Distilled variants prioritize speed with minimal quality loss, achieving near real-time performance. Available in 4B and 9B parameter sizes, these models support text-to-image and image editing processes, accommodating single-reference and multi-reference workflows. Designed for latency-critical applications, the models enable creative iteration with end-to-end inference times as low as one second, making them suitable for live previews and interactive applications. Users can explore these models on platforms like ComfyUI and HuggingFace, with the Distilled models particularly suited for production deployments and real-time applications.

ComfyUI introduces a set of preprocessor-focused template workflows designed to enhance the consistency and reusability of common tasks in image, animation, and video workflows, including depth estimation, lineart conversion, pose detection, normals estimation, and frame interpolation. These modular workflows allow for faster iteration, easier debugging, and more predictable results by cleanly separating preprocessing from generation processes. Depth estimation provides spatial awareness for edits and relighting, while lineart conversion preserves structural guidance without overconstraining style. Pose detection extracts keypoints for control over posture and movement, and normals estimation offers detailed surface orientation for relighting and stylization. Frame interpolation smooths motion by generating intermediate frames, improving temporal consistency without regenerating source frames. These workflows can be integrated into larger ComfyUI graphs, providing reliable building blocks for creative projects.

Kling 2.6 Motion Control, now available in ComfyUI, allows users to achieve precise control over character actions and expressions by transferring motion from a reference video to a character image. This update supports full-body motion transfer with accurate timing and can handle complex actions such as dancing, skating, and martial arts, ensuring strong consistency and natural movement. Users can control scenes, clothing, and environments through prompts, and the feature offers flexible orientation modes for character-facing or camera-driven motion. To utilize this feature, users must update ComfyUI to version 0.7.0 or above and prepare a reference image and video for the workflow. Best practices suggest matching character proportions, selecting quality motion references, and ensuring the character's visibility for optimal results.

Between October 17th and 19th, 2024, a malicious actor uploaded two node packs containing malware to the Comfy Registry, which were subsequently downloaded 790 times before being banned. The Registry team, prompted by an automated security scanner and a manual review, banned the malicious versions on October 21st, 2025. During this period, a third node pack with a similar name was created by a different user. On October 21st, an internal security scanner identified suspicious code patterns in the initial node packs, and the registry maintainer confirmed the presence of malware, leading to the ban of all affected versions from the Comfy Node Registry.

In 2025, ComfyUI has evolved from an experimental tool into a robust, industry-grade creative engine utilized in various sectors such as film, VFX, animation, and gaming, becoming a trusted resource for leading artists and production teams worldwide. Over the past year, ComfyUI's capabilities have been showcased on significant platforms and integrated into demanding production workflows, demonstrating its reliability, control, and adaptability. The community behind ComfyUI, characterized by collaboration and shared learning, has contributed to its growth, pushing creative boundaries and tackling cutting-edge challenges. Grounded in open-source principles and a scalable, node-based system, ComfyUI aims to empower creators by offering a controllable environment for experimentation and innovation, inviting those passionate about creativity, craftsmanship, and problem-solving to join in shaping its future.

ComfyUI has introduced several optimizations for NVIDIA GPUs, including the NVFP4 quantization format for Blackwell GPUs, async offloading, and pinned memory, offering performance enhancements without hardware upgrades. The NVFP4 quantization format utilizes the FP4 hardware on NVIDIA's Blackwell architecture to potentially double performance on RTX 50-series or Blackwell Pro GPUs, provided PyTorch is built with CUDA 13.0. Meanwhile, async offloading and pinned memory, enabled by default for all NVIDIA GPUs, can improve sampling speed by 10-50% depending on the hardware and model setup, particularly benefiting scenarios where model weights cannot fully fit in VRAM. These optimizations are contingent on the PCIe generation and lane count, with greater improvements seen in PCIe 5.0 compared to PCIe 4.0. As RAM prices rise, ComfyUI is also working on RAM-usage optimizations and encourages users to engage with their community for further updates and testing opportunities.

LTX-2, an open-source audio-video AI model, is now integrated into ComfyUI, offering high-quality visual output with efficient resource and speed usage. It allows for synchronized generation of motion, dialogue, background noise, and music in a single pass, creating cohesive audio-video experiences. Developers benefit from a customizable and transparent framework, enabling creative freedom and control. The model supports consumer-grade hardware and features such as video-to-video control and keyframe-driven generation, with capabilities for native upscaling and prompt enhancement. LTX-2's integration with NVIDIA optimizations allows for cloud-class 4K video production locally at increased speeds and reduced VRAM usage.

Official support for AMD ROCm™ has been introduced to the ComfyUI Desktop app on Windows, starting with version v0.7.0, enabling users to leverage the full capabilities of AMD Radeon™ GPUs and Ryzen™ AI processors for enhanced performance. This update allows ComfyUI to operate seamlessly across Windows Desktop, Git, and Portable versions, with automatic selection of AMD ROCm™ during installation. The release is based on ROCm 7.1.1 and suggests using the AMD ROCm™ 7.1.1 Preview driver for optimal performance, including support for the --use-pytorch-cross-attention flag, which is anticipated to be integrated into an upcoming AMD Software: Adrenalin™ Edition driver release. Future updates are expected to include further performance enhancements, expanded support, and bug fixes.