October 2024 Summaries

Stable Diffusion 3.5 Medium, introduced by Stability AI, is now supported in ComfyUI and is designed to operate on consumer GPUs with 2.6 billion parameters, supporting resolutions from 0.25 to 2 megapixels. This model can be utilized through updated Example Workflows, requiring users to update ComfyUI, download the model to their checkpoint folder, and ensure necessary files like clip_g.safetensors are in place. For those with low RAM configurations, alternatives include using the fp8_scaled workflow and dual-clip options. To simplify the process for users, an all-in-one packaged checkpoint has been provided, allowing for a straightforward "Image Generation" workflow without extensive configuration.

Starting on November 15, 2024, ComfyUI will make its new Beta UI the default interface, following extensive testing and development. This transition aims to eventually phase out the legacy UI to streamline development and reduce maintenance for both core and extension developers. Users can immediately switch to the new UI through the settings menu, with the option to revert to the legacy version temporarily if needed. While existing workflows will remain unaffected, some frontend extensions may not yet be fully compatible, prompting users to file issues on GitHub if they encounter problems. The transition will be informed by user feedback and adoption rates, ensuring a smooth changeover. Extension developers are encouraged to update and test their extensions for compatibility with the new UI, and additional technical documentation will be provided. Users encountering issues with the new UI are advised to report bugs, seek support via Discord, and provide feedback to facilitate improvements.

ComfyUI has announced support for the newly released Stable Diffusion 3.5 models, which include Stable Diffusion 3.5 Large with 8 billion parameters for high-quality image generation, and the faster Stable Diffusion 3.5 Large Turbo. Set for release on October 29th, the Stable Diffusion 3.5 Medium model caters to consumer hardware with its 2.6 billion parameters. These models are available for both commercial and non-commercial use under the Stability AI Community License. ComfyUI users can update to the latest version to leverage these models via node-based workflows, with options for low RAM environments. The update facilitates ease of use through auto-download features and workflow examples, aiming to foster community experimentation and feedback. Additionally, ComfyUI is preparing to release Stable Diffusion 3.5 Medium with enhanced control features aimed at professional applications.

ComfyUI has launched its V1 release, introducing a range of enhancements aimed at improving user experience and productivity. The update includes a fully packaged desktop version that is code-signed for security, offers cross-platform compatibility, automatic updates, and a lightweight package. A recommended Python environment is included to ease installation, and the ComfyUI Manager is now shipped by default for seamless access to the latest nodes. The new user interface features a top menu bar, easier access to models and logs, a model library, and a workflow browser, allowing users to efficiently manage workflows and download models automatically. ComfyUI has also developed a registry for custom nodes akin to NPM, with plans for security scanning and improved dependency resolution. While the desktop app is in closed beta, an open beta and the open sourcing of the code are planned soon. The new V1 UI will become the standard, as the original UI enters maintenance mode, encouraging developers to support the updated version.