TT CLIP Text Encode FLUX (Context)

Most text-encoder nodes are where you type your prompt. This one is where you don't. TT CLIP Text Encode FLUX (Context) takes a single TT_CONTEXT in and produces POSITIVE and NEGATIVE conditioning - the prompts come from the workflow config that's already riding inside the context. It's the FLUX encoder for people who've decided their prompts and settings live in one place, up top, and don't want to retype them in the middle of a graph.

It's the context-flavored sibling of TT CLIP Text Encode FLUX, and it exists for the same reason the whole context system exists: fewer wires, one source of truth. The workflow-settings node holds clip_l_positive, t5xxl_positive, the negatives, and guidance; this node quietly reads them when the context arrives.

How it works

The encoding itself is identical to the non-context version - CLIP-L tokens and T5-XXL tokens get tokenized separately and combined, then encoded with the guidance value attached. It pulls clip, the four prompt strings, and guidance out of the context's workflow config, builds the conditioning, stores it back in the context, and returns three outputs:

• CONTEXT - updated context, now carrying the conditioning for anything downstream.
• POSITIVE / NEGATIVE - the plain conditioning sockets, which you can wire to any sampler (TenserTensor's or stock).

Why this node, in practice

Two reasons people reach for it. First, the prompts live exactly once, in the workflow-settings node - when you're iterating on a FLUX prompt you change it in one place and every branch of a big graph picks it up. Second, it keeps the graph compact: the encoder doesn't sprawl with text fields. This is the same "state down one wire" philosophy as rgthree's Context nodes, and the same trade - TenserTensor's TT_CONTEXT is a proprietary type that only connects to other TenserTensor nodes, so you're committing to the pack's context style.

Caveats and install

The error you'll actually hit: if the workflow config is missing the prompt fields - usually because you forgot the workflow-settings node, or wired the context from a node that didn't carry a config - this node raises a clear error asking for its config. Trace the context back to its source before suspecting the node.

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

or ComfyUI Manager → "TenserTensor" → install → restart. Deps: gguf, kornia.

And the standing pack note: it's a legacy V1 node, in Deprecated/ since the pack's migration to ComfyUI's API V3, with removal planned for a future major release. It works today. Also remember the FLUX guidance reality that applies here too - the negative prompts it encodes are mostly inert on guidance-distilled FLUX, so write positives as presence and keep an eye on the guidance value in your config.