InvokeAI vs tl;dv

6.14.0 is now final after two release candidates dating back to 31 July. The shipped cut carries video generation via Wan 2.2 — text-to-video, image-to-video and interpolation between two stills, driven from the workflow editor with template workflows for concatenating longer productions — alongside multi-GPU execution, native Intel XPU support, and a wide model list: Krea.2 Turbo and Raw, Flux.2 Dev, Flux.2 PiD super-resolution to 4K, Ernie Turbo, Ideogram 4, Anima controlnets and FP8 across several families.

InvokeAI broadened on two axes at once and has now landed both. What it can generate crossed from images into video, which is the larger change; what it can run on grew from multi-GPU in RC1 to native Intel XPU in RC2 and ROCm 7.1 in the preceding maintenance cut. With 6.14.0 out, the model-family additions are the routine work and the hardware breadth is the differentiator — an image tool that runs on Intel and AMD as well as NVIDIA and now produces video is competing on reach rather than on any single model.

Expect the next cycle to fill in what RC notes flagged as staged — LoRA, ControlNet and IPAdapter coverage for the newly added text-to-image models — and further video work now that Wan 2.2 is in stable. Whether video moves out of the workflow editor into the main canvas isn't indicated in these entries.