ComfyUI v0.24.0 adds Ideogram 4 support, fixes MultiGPU freezes

ComfyUI v0.24.0 is a feature-packed update to the popular Stable Diffusion workflow UI, bringing significant new model support and stability improvements. The headline addition is the Ideogram 4 model, delivered through a new Partner Node (feat: Ideogram V4 node) and core model support. Users can now leverage Ideogram's latest text-to-image capabilities directly within ComfyUI workflows. Additionally, the PiD (Prompt Image Diffusion) node now supports SDXL and QwenImage, expanding its versatility. The release also updates comfy-aimdo to 0.4.8 and includes a Radiance variant with nonzero txt_ids support.

On the bug-fix and performance front, several critical issues were resolved. A longstanding freeze bug in MultiGPU CFG Split when manually aborting has been fixed by new contributor quasar-of-mikus. The Math Expression node no longer crashes on large integer results thanks to vidigoat's OverflowError fix. Memory usage for Ideogram 4 on non-dynamic VRAM systems is now optimized. Other fixes include cast/dtype adjustments for birefnet and dino3 models, a fix for triposplat preview with old offloading, and reverting partner node category changes. The release also removes the old no-comfy-kitchen fallback and marks DualModelGuider as experimental for future development. With 5 new contributors and 14 merged PRs, v0.24.0 strengthens ComfyUI as a production-ready platform for advanced AI image generation.