ComfyUI continues to be the go-to node-based interface for generative image and video workflows, and its release pace in late October and November 2026 was brisk. Here is what stood out from v0.43 through v0.45.2.
Video and audio partner nodes
Version 0.45.0, released on November 10, added several partner integrations: Gemini Omni Flash 1.2 in the Omni video node, Pika H3 Flex nodes for video and reference-to-video, a set of Pika audio nodes covering soundtrack, sound effects and music, and a node for Tencent’s HY language model. A follow-up, 0.45.1, brought LTX VFX tools including alpha-channel output, SDR-to-HDR conversion, reframing and upscaling.
Professional output and efficiency
Version 0.44.0, on November 4, added ProRes video saving with an option for chroma subsampling, a welcome feature for editors who want to move generated clips into conventional post-production pipelines. The same release made the fp8_e4m3fn_fast path work with dynamic VRAM instead of falling back to bf16 computation. In 0.45.0, the Get Video Components node now keeps decoded frames only once instead of twice, trimming memory use.
Stability and the wider ecosystem
Version 0.45.2, on November 12, fixed a problem where some workflows did not load correctly. Elsewhere, ByteDance’s Seedream 5.5 image model, released on November 9, runs inside ComfyUI through the partner API, which means no model download is needed but usage is billed to your Comfy API account. Community nodes are growing too, including a ShotGrid integration that publishes results with a production-tracking record.
Because partner nodes depend on core nodes, update ComfyUI before loading a new workflow, and use the nightly version if nodes appear to be missing.