🔶 Main_Prompt Clip Endcode

Think of this as the "unwrap and execute" step for a chaosaiart_Prompt bundle - it's the node that takes the packaged MAIN_PROMPT (positive text, negative text, and whatever LoRA chain got attached) and actually turns it into the MODEL and CONDITIONING your KSampler needs. In a standard ComfyUI workflow this would be a LoraLoader plus two separate CLIPTextEncode nodes; here it's one node doing all three jobs at once because the data arrives pre-bundled.

The required inputs are simple: model and clip, straight from your checkpoint loader, and main_prompt (type MAIN_PROMPT), which has to come from a chaosaiart_Prompt node - or, in a frame-by-frame animation setup, from wherever your chaosaiart_Prompt_mixer_byFrame node's resolved output ends up feeding this. This node doesn't have its own text fields; all the actual prompt writing happens upstream.

What it does internally: if the MAIN_PROMPT bundle it received has a LoRA chain attached (via the add_lora input back on the Prompt node), this node applies that chain to both the model and the CLIP encoder - which is exactly why it re-outputs MODEL rather than just passing your input model through unchanged. If there's no LoRA in the bundle, the model output is effectively a passthrough. It then encodes the positive and negative text through the given CLIP, producing standard POSITIV and NEGATIV conditioning outputs that plug straight into any KSampler - this pack's own, or a stock one.

That last point is worth underlining: even though this node lives inside the Chaosaiart ecosystem and consumes a Chaosaiart-only type on the way in, everything it produces - MODEL, POSITIV, NEGATIV - is completely standard ComfyUI typing. This is the seam where the pack's custom prompt system hands back off to the rest of your graph, so you're free to route its outputs into a plain KSampler if you're not using the pack's animation sampler at all.

Install: search "Chaosaiart-Nodes" in ComfyUI Manager, or use "Install via Git URL" with https://github.com/chaosaiart/Chaosaiart-Nodes. Manually: cd ComfyUI/custom_nodes && git clone https://github.com/chaosaiart/Chaosaiart-Nodes, then restart ComfyUI. On Linux, also run pip install opencv-python and pip install tqdm inside your venv.

The one thing to double-check if your LoRA doesn't seem to be applying: this node only picks up a LoRA that was actually wired into the add_lora chain on the upstream chaosaiart_Prompt (or chaosaiart_Prompt_Frame) node. If you built a chaosaiart_lora node but forgot to connect its LORA output anywhere, this node has nothing to apply - there's no error, your model just comes out unmodified.

It's also worth being deliberate about where this sits in a frame-by-frame graph: because it re-encodes text and re-applies LoRA weights fresh every single run, it's meant to sit downstream of whatever resolves your current prompt (the mixer, if you're using per-frame overrides) rather than being wired once to a single static chaosaiart_Prompt and forgotten. If your prompt is supposed to be changing across an animation and the output never seems to shift, check that this node is actually receiving the mixer's resolved bundle each run and not a fixed one that bypassed the frame logic entirely.