RM Negative Text Embed

The stock CLIP Text Encode (Negative Prompt) is fine - until your negative prompt has two sources: a base negative that lives in a styles file, and the extra stuff you want to type in for this particular run. Then you're juggling a concatenation node and a separator string and praying about commas. This node bakes that merge into the encoder itself: it takes a string_input and a text field, joins them, and encodes the result straight to negative conditioning in one step.

How it works

Both text sources get trimmed, joined with ", ", tokenized through the CLIP you feed in, and encoded with encode_from_tokens_scheduled - which is the same scheduled-encoding path the stock encoders use, so dynamicPrompts-style syntax in the text field still behaves normally. If both sources are empty, it encodes an empty string rather than erroring. The only hard failure is a None clip, which raises a clear "clip input is invalid" error.

That's the whole mechanism, and it's the right size for the job. You're not getting weighted unions or masking - this is a convenience node that removes one fiddly junction from every workflow that already has a style pipe or a base-prompt node upstream.

Inputs and outputs that matter

• clip (CLIP, required) - your text encoder, normally straight from the checkpoint/CLIP loader.
• text (STRING, multiline, required) - your hand-typed negative.
• string_input (STRING, optional) - the external source: connect the Negative output of RM Styles Full, or any node emitting a string.
• Output: Negative (CONDITIONING) - straight into your KSampler's negative input.

If both inputs carry a negative, the style string goes first and your manual text second - worth remembering for prompt-weight interplay, since earlier tags can set the tone for later ones.

How to install it

Part of the RMAutomation pack:

cd ComfyUI/custom_nodes
git clone https://github.com/Moser9815/ComfyUI-RMAutomation

Restart ComfyUI and it's under RMAutomation/Text. It's pure CLIP work - no extra dependencies.

Common issues

The order of operations is the only real trap: string_input is prepended, so if your styles file already includes a generic negative and you also type "bad hands" into the text box, the string wins the front position. That's usually what you want, but it surprises people who expected the manual text to lead. Beyond that, this node is as boring as it should be - wire the clip, type your text, forget it. The sibling node RM Positive Text Embed is the exact mirror for the positive side if you want the same merge behavior there.