Hypnodes Prompt Encoder

The Prompt Encoder is the pack's "base conditioning" node - the one that encodes the global positive and negative prompts from the Control Hub into real CLIP conditioning you can feed a sampler. What makes it distinctive is what it doesn't have: no text fields at all. Everything comes from the hn_config bus.

What it takes

Just two inputs:

• clip - from your checkpoint or CLIP loader.
• hn_config - the Control Hub's bus.

The source calls it "The Tiny Encoder," which is an accurate self-assessment. It reads global_prompt_start (the Hub's global positive) and global_negative_prompt, runs a standard CLIPTextEncode on each, and hands them out.

The outputs

• positive - CONDITIONING for the global positive prompt.
• negative - CONDITIONING for the global negative prompt.

Wire both into your sampler (via a KSampler, or into a guider if you're using a custom sampler stack). This is the conditioning that represents the whole scene - environment, lighting, quality tags, and your universal negatives - as opposed to the per-character conditionings from the Hypnodes Conditioner.

Where it fits in the workflow

The division of labor is clean and worth internalizing:

• Hypnodes Prompt Encoder - the global/base pass. Scene, quality, negatives for the whole image.
• Hypnodes Conditioner - the per-character pass. Each character's regional conditioning, paired with a mask.

If you're running a regional workflow with an Attention Coupler or Regional Prompter, the Prompt Encoder's positive is what those tools use as the "base" layer that regions sit on top of, and the negative is what applies everywhere. If you're skipping regional prompting entirely, this node alone gives you a normal text-to-image encode that just happens to be driven by the Hub's text boxes.

The catch to know

Because it has no text fields, it's useless without the Control Hub - the hn_config bus is its only source of text. If you're not adopting the Hypnodes system and just want a convenient encoder, a stock CLIP Text Encode node does the same job. This node is only worth reaching for inside a Hypnodes workflow, where it earns its keep by keeping your global prompts in one editable place.

Install

ComfyUI Manager (search Hypnodes) or:

cd ComfyUI/custom_nodes
git clone https://github.com/hypnichorse/ComfyUI-Hypnodes

Restart ComfyUI. No model downloads; the pack's only dependencies are numpy and pillow, both already in ComfyUI.