May 2025 Summaries

FLUX.1 Kontext has been introduced as an official ComfyUI Partner Node, offering a suite of generative flow matching models focused on text-and-image-driven editing and generation. Unlike conventional text-to-image models, Kontext enables in-context image manipulation using both text and images, supporting precise object modification, character-consistent multi-step editing, style preservation, and transfer. The two models in the series, Kontext Pro and Kontext Max, cater to different needs, with Pro ideal for editing, composing, and remixing, and Max specializing in typography and prompt precision. Users can access these features by updating ComfyUI, purchasing credits, and utilizing the Flux.1 Kontext templates available in the Image API section, allowing for ultra-fast inference and the ability to control aspects like framing, camera angle, and pose.

The Comfy Bounty Program is a new initiative designed to expand and enhance the ComfyUI ecosystem by engaging developers, designers, testers, and creative contributors through a task-based reward system. Participants can claim tasks from a public Bounty Task Table that includes descriptions, category labels, estimated bounty amounts, GitHub links, and current status, allowing them to earn money by completing projects, fixing bugs, or developing features. The program aims to leverage the open-source community to address potential improvements for ComfyUI while also identifying talented individuals for possible recruitment. To participate, contributors must review available tasks, submit a proposal outlining their approach and timeline, and wait for approval before proceeding, after which they can complete the task and receive their bounty.

ComfyUI Partner Nodes Wave 2 has been launched, introducing six new models that seamlessly integrate 3D asset generation, language model processing, and media compositing within the ComfyUI platform. This update supports large language models such as Google's Gemini 2.5 and OpenAI's GPT-4.1, as well as image and video models from RunwayML and Vidu. The new features enable users to generate text, build logic branches, and create complex workflows using language models directly in the interface. The platform's upgraded 3D support includes tools for geometry and texture workflows from Hyper3D AI and Tripo AI, allowing for high-quality mesh creation with detailed presets. Additionally, users can mix and match online APIs with open-source models to create customized pipelines, enhancing both prototyping and production processes. The update aims to streamline creative workflows by integrating various AI models and tools into an easy-to-use node-based system.

ComfyUI has announced native support for Wan2.1-VACE, an open-source video editing model developed by the Alibaba Tongyi team, which enhances video creation through its multi-tasking capabilities, high-resolution processing, and flexible multi-modal inputs. This model integrates diverse video editing functions such as Move-Anything, Swap-Anything, and Animate-Anything within a single framework, streamlining workflows for users. Additionally, ComfyUI has refined its Ace-Step Music Generation Workflow, introducing a Tonemap Multiplier feature to adjust vocal volume, which was showcased in a recent video. Users are encouraged to update their ComfyUI or ComfyUI Desktop, access templates, download models, and explore workflows such as VACE Text-to-Video, Image-to-Video, and Video-to-Video, while further details and documentation are available for those interested in deeper exploration and creation.

ComfyUI has addressed previous authentication challenges associated with Partner Nodes by introducing the ability to log in using ComfyUI API Keys, alongside the existing whitelisted domain method. This update, announced shortly after the initial launch, enhances accessibility by allowing users to operate ComfyUI in a head-less mode and simplifies the login process. Users are encouraged to update to the latest stable version of ComfyUI or its Desktop variant, create an account on the ComfyUI platform, and add their API key through the settings to take advantage of this new feature. Further details are available in the ComfyUI documentation, with the blog encouraging readers to subscribe for ongoing updates.

ComfyUI has announced native support for the Ace-Step audio model, an open-source music generation tool developed by ACE Studio and StepFun that can produce a variety of music genres and supports multiple languages. The model, considered a significant advancement in audio generation, allows extensive customization through techniques like LoRA and ControlNet and is available for commercial use under the Apache-2.0 license, offering fast synthesis times with the capability of generating up to four minutes of music in 20 seconds on an A100 GPU. Additionally, ComfyUI now supports HiDream-E1, an open-source interactive image editing model by HiDream-ai, and has introduced an FP8 version of the Wan2.1 FLF2V model, enhancing performance with reduced VRAM requirements. These updates, designed to enhance user experience and creativity, are available for download and integration within ComfyUI, with further details accessible in the provided documentation.

ComfyUI has introduced Partner Nodes, which are integrated nodes within its platform that allow for the use of paid model APIs, expanding its creative automation capabilities. Recently, the platform launched GPT-Image-1 as a Partner Node and, following user feedback, has added support for ten more model families and 62 new nodes, including models like Black Forest Labs Flux, Stability AI Stable Image Ultra, and OpenAI GPT4o image. Users can access these features by updating ComfyUI, logging in, purchasing credits, and using templates from the Image or Video API sections. The platform remains free and open-source, with the use of Partner Nodes being optional. Additionally, ComfyUI has introduced a new branding look that includes a fresh logo and color scheme, reflecting its evolution into a robust and enduring creative tool, while maintaining its playful origins.