🔶 Lora Advanced +add_lora

Same job as chaosaiart_lora, with two additions aimed specifically at the pack's frame-by-frame animation use case: routing control over which prompt a LoRA's trigger words attach to, and live-overridable strength values you can drive from elsewhere in your graph instead of hardcoding a fixed number on the node.

The base fields are identical to the plain Lora node: lora_name (a dropdown built from your LoRA folder at startup - restart ComfyUI after adding new files), strength_model and strength_clip (default 1, range -20 to 20 - most LoRAs actually want somewhere around 0.5 to 0.8 rather than full strength before they start overpowering your checkpoint), and the optional add_lora input for chaining another Chaosaiart LoRA node into a stack.

What's new here is lora_type, a required enum choosing Positiv_Prompt or Negativ_Prompt - this decides whether the trigger-word metadata this node reads off your LoRA file (the same idrirap-inspired feature the plain Lora node has) gets treated as belonging to your positive or negative prompt. That matters for LoRAs whose intended use is closer to a negative embedding than a style adapter - you want their tags steering the prompt they're actually meant for, not defaulting to positive every time.

The other addition is the pair of optional overrides: strength_model_override and strength_clip_override (both FLOAT, unbounded by a fixed range since they're meant to be driven dynamically). Wire a chaosaiart_Number2 node - or any other float source elsewhere in your graph - into these, and the override value wins over whatever's set on strength_model/strength_clip. This is built for the pack's animation loop specifically: you can ramp a LoRA's strength up or down across a sequence of frames by feeding these overrides from something frame-aware, instead of manually editing the node's fixed value between renders.

Outputs are the same shape as the plain Lora node: LORA (the chain, feeding into another LoRA node's add_lora or into chaosaiart_Prompt/chaosaiart_Prompt_Frame's add_lora) and INFO (the trigger-word text pulled from metadata).

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. On Linux, also pip install opencv-python and pip install tqdm inside your venv.

If you're not actually using the override inputs or the positive/negative routing, the plain chaosaiart_lora node does the same core job with one less field to think about - reach for this one specifically when you need strength that changes across your animation, or a LoRA whose trigger words genuinely belong in the negative prompt.

One thing to keep in mind with the override pair: since strength_model_override and strength_clip_override are optional and unbounded, nothing on this node stops you from leaving one connected and one not, or feeding in a value way outside the sane -20 to 20 range the base fields enforce. If a LoRA suddenly looks wildly over- or under-baked after you wire in an override source, that's the first thing worth checking - the override silently wins over the widget value, so a stray or forgotten connection is easy to miss compared to a value you can see sitting right there in the node.