ComfyUI v0.44.0 Released with Expanded Partner Nodes

ComfyUI v0.44.0 Released with Expanded Partner Nodes

At a Glance

Item Value
Repository Comfy-Org/ComfyUI
Version v0.44.0
Published 2026-11-04
License GPL-3.0
Source type Primary source (the publisher itself)

Values determined by this site’s code when the information was collected. Dates are JST.

Overview

Comfy-Org/ComfyUI has released its latest version, v0.44.0. ComfyUI is a powerful, modular diffusion model GUI, API, and backend featuring a graph/node interface. This update significantly expands “Partner Nodes" related to external services such as MiniMax, OpenAI, and Topaz, greatly broadening the options for video generation, upscaling, and utilizing specific external models.

Breaking Changes and Deprecations

In this release, the following node has been deprecated:

Old New
Seedance 1.x video nodes Deprecated

Users building workflows with Seedance 1.x video nodes may need to consider alternative methods for future compatibility.

Key Changes

MiniMax-Related Enhancements

Extensive support related to MiniMax-H3 has been added. Specifically, this includes support for the MiniMax-H3 latent upscaler model, MiniMax Sol-H3-VAE, and the MiniMax-H3 context window. Additionally, various nodes such as H3 Max Recast, Extend, Reframe, Insert, Lip Sync, and Camera Control have been added. This affects users who want to use MiniMax models for advanced control such as video reconstruction, extension, and lip-syncing. To use these features, add the new nodes to your workflow.

Video Generation and Upscaling Improvements

Numerous features to enhance video generation workflows have been introduced. Targets for 1080p, 2K, and 4K have been added to BFL’s Flux Video Upscale V2 node, alongside the introduction of Pruna’s P-Video Upscale node. Furthermore, users can now select the ProRes format and different chroma subsampling options when saving videos. This affects users seeking high-resolution video generation or specific video quality. You can use these new targets and codec settings within your workflows.

Expansion of External Models and Features via Partner Nodes

“Partner Nodes," which support various external services and models, have been significantly expanded. This includes OpenAI’s GPT-6.1 Terra and Sol models, the addition of the Wonder 4 model to Topaz Image Enhance, context window regional editing in Ideogram 4.5 Precise Edit, Sonilo Music 1.5 nodes, Patina texture nodes, and VEED Dubbing nodes. This affects users who want to directly operate these external models and features within the graph. Add the new nodes to your workflow to use them.

VAE and Computational Precision Optimization

A new node has been added to explicitly set the compute dtype for VAE calculations. Accompanying this, a crash related to VAE Compute Dtype that occurred in the LTX 2.5 diffusion VAE decoder has been fixed. Additionally, an issue where fp8_e4m3fn_fast fell back to bf16 calculations during dynamic VRAM usage was resolved. This improves stability for users working under VRAM constraints who want to use specific precisions, as well as those using LTX 2.5. Use the new node to explicitly specify the data type if necessary.

Supported Models and Hardware

With this update, support for the following models and features has been expanded:

  • MiniMax-H3: Support added for latent upscaler models, Sol-H3-VAE, and context windows.
  • Flux Video Upscale V2: Target resolutions for 1080p, 2K, and 4K added.
  • Topaz Image Enhance: Wonder 4 model added.
  • Ideogram 4.5 Precise Edit: Regional editing via context window is available.
  • OpenAI: Support added for GPT-6.1 Terra and Sol models.
  • PH3Image: Newly supported.
  • Sonilo Music 1.5: New nodes added.
  • Patina: Texture nodes added.
  • VEED: Dubbing nodes added.
  • Pruna: P-Video Upscale nodes added.

How to Get It

If you installed manually, update your repository to the latest state. To install manager dependencies, run the following command:

pip install -r manager_requirements.txt

To enable the manager when running ComfyUI, execute it with the following flag:

python main.py --enable-manager

Related Articles

What to Read Next

Sources