TT Context Set Image

Every context workflow eventually hits the same gap: you've got an image in hand - a loaded reference, a ControlNet input, a preview from an earlier pass - and the downstream context node wants that image inside the context. TT Context Set Image is the two-input node that does it: context and image in, updated CONTEXT out, with the image stored in the context for whatever comes next.

It's the pixel-space sibling of TT Context Set Latent, and it belongs to TenserTensor's context system - the "bundle state, pass it down one wire" pattern that keeps big graphs from becoming wire spaghetti. The setter nodes are the punctuation marks: they're how state that gets produced off to the side ends up back in the main flow.

What it's for

The concrete cases:

• Reference-image workflows. You load a source image, want to carry it through the graph for img2img or as a visual reference, and a context-aware encode node expects to pull image out of the context. This node is the handoff.
• Inpainting / masking setups. You've got an image and a mask; this stores the image so the context has it when a downstream node asks.
• Multi-pass previews. A sampler produced a latent, you decoded and tweaked it, and now the revised image should replace what the context was carrying.

That last point matters: the node overwrites the image field. If your context already had an image and you wire a new one in, the old one is gone. That's usually what you want - a context should hold the current state, not a history - but it's the first thing people trip on.

Caveats

Same rules as the rest of the pack's context family: TT_CONTEXT is proprietary, so this node only talks to other TenserTensor context nodes, and it's a legacy V1 node sitting in Deprecated/ since the pack's migration to ComfyUI's API V3 (removal planned for a future major release). It works today.

The other gotcha: it returns only the context - the image is not forwarded as a separate output. If something downstream needs the plain IMAGE socket too, keep a separate wire from the node that produced it. Beginners assume a setter forwards its input; it doesn't, by design.

Install

cd ComfyUI/custom_nodes
git clone https://github.com/tenser-tensor/ComfyUI-TenserTensor

or ComfyUI Manager → search "TenserTensor" → install → restart. Deps: gguf and kornia, handled automatically by Manager.