ComfyUI v0.44.0 Release: New Models and ProRes Export

The stable version of ComfyUI v0.44.0 has been released, including 40 commits in the core and a significant update to partner nodes for advanced generation tasks.

Apple ProRes Video Export

Support for MOV format with ProRes codec (prores_ks) has been added to the Save Video node. It uses FFmpeg's software encoder without paid plugins. Color subsampling modes 4:4:4, 4:2:2, 4:2:0, and Auto are available, as well as ProRes HQ and 4444 profiles.

New Model Integration

The core now natively supports P-H3-Image from Pruna AI for direct loading and inference. Built-in support for MiniMax Sol-H3-VAE and the MiniMax-H3 latent upscaler with extended context windows for video pipelines has been implemented.

Partner extensions include:

• Flux Video Upscale V2 by Black Forest Labs with output in 1080p, 2K, and 4K
• Official nodes for OpenAI GPT-6.1 Terra and Sol models
• Sonilo Music 1.5 music generation node
• VEED dubbing node
• Wonder 4 model in Topaz Image Enhance
• P-Video Upscale by Pruna

Technical Improvements and Interface

Python bypass for memfd_create introduced, redundant torch.equal calls eliminated. Fixed fp8_e4m3fn_fast crash in bf16 during dynamic VRAM management. The web interface updated to version 1.58.1, workflow template library to v0.11.101.

Update is available via the built-in ComfyUI Manager or git pull.