v0.19.1

Comfy-Org has rolled out version 0.19.1 of ComfyUI, a widely-used visual programming interface for building and running complex AI workflows, particularly for Stable Diffusion. This incremental release focuses on bug fixes and feature integration, addressing specific issues reported by its community of over 109,000 GitHub stars. The update patches problems with the Ernie image model on hardware lacking full fp64 (double-precision) support and ensures the Ministral text generation model works correctly within the node-based system.

A notable addition is the integration of Sonilo Audio nodes, contributed by community member bigcat88. This expands ComfyUI's capabilities beyond visual generation into the audio domain, allowing users to incorporate sound generation or processing into their automated pipelines. Other tweaks include updates to workflow templates and a fix for the preview text node to properly output strings. The release follows the typical ComfyUI pattern of rapid, community-driven iteration, with just three commits separating it from the previous version, v0.19.0.