Ultimate Lora Loader (Civitai):
UltimateLoraLoaderCivitai applies a stack of LoRAs to a ComfyUI model. Its frontend manages the stack as rows that you can add, remove, reorder, enable or disable, and adjust for strength; the node then applies enabled LoRAs in their listed order. This lets you combine visual effects in one place while controlling how strongly each affects the image model and, when available, its text-conditioning component. Despite its name, the supplied code does not show a Civitai download or search feature: it loads LoRA files that ComfyUI can already locate locally.
Ultimate Lora Loader (Civitai) Input Parameters:
model
The required MODEL connection supplies the image-generation model that the LoRAs will modify. Each applicable LoRA is applied to the result of the preceding one, so the stack's order matters. Connect a model from an earlier loader or processing node. This socket has no numeric minimum, maximum, or default in the supplied interface; a MODEL connection is required.
loras_data
This required STRING widget stores the LoRA stack as JSON text, although the frontend is designed to manage it through its row-based interface. Its default is [], an empty stack that leaves the supplied model unchanged. Each entry can identify a local LoRA through lora, control whether it is used through enabled (default true), and set separate strengthModel and strengthClip values; either strength falls back to strength and then to 1.0 when absent. Entries are processed in array order. An entry without a lora name is skipped, as is one whose effective model and CLIP strengths are both zero. The interface specifies no minimum or maximum string length and no numeric strength limits. Invalid JSON is silently treated as an empty stack, so preserving valid JSON matters if you edit this value directly.
clip
The optional CLIP connection supplies the text-conditioning component for LoRAs to modify alongside the image model. Connect it when you want the stack's CLIP strengths to take effect. When it is unconnected, the node skips CLIP patching by using a CLIP strength of zero for every entry, regardless of the stored setting. The supplied interface defines no numeric minimum or maximum; its effective default is no CLIP connection.
Ultimate Lora Loader (Civitai) Output Parameters:
MODEL
The first output is the MODEL after the node has applied each eligible LoRA in stack order. Connect it to the next node that needs the modified image model. If the stack is empty, its JSON cannot be parsed, or every entry is skipped, this output remains the supplied model.
CLIP
The second output is the CLIP after the node has applied the eligible LoRAs' CLIP strengths. Use it downstream where a modified text-conditioning component is needed. If no CLIP was connected, this output is None; if one was connected but no LoRA applied a CLIP change, it remains the supplied CLIP.
Ultimate Lora Loader (Civitai) Usage Tips:
- Add LoRAs through the node's frontend controls when available. The controls keep the JSON stack synchronized and are easier to manage than editing
loras_databy hand. - Reorder entries to control application order: each enabled LoRA receives the model and CLIP produced by the previous entry.
- Disable an entry to compare a result without removing its settings. An entry with both effective strengths set to zero is also skipped.
- Connect
clipwhen you need a LoRA's CLIP effect. For a model-only workflow, leave it unconnected; stored CLIP strengths will then have no effect. - Check that each selected LoRA is available in ComfyUI's configured LoRA locations. A name that cannot be resolved is logged and skipped, while the remaining entries continue.
Ultimate Lora Loader (Civitai) Common Errors and Solutions:
Could not find lora: <name>, skipping.
- Explanation: The node could not resolve the named file in ComfyUI's configured LoRA locations. The actual console line also begins with the node's configured log prefix; the missing entry is skipped rather than stopping the whole stack.
- Solution: Check that the file is present in a configured LoRA location, then select its correct name in the node's frontend.
No LoRA effect and no error after editing loras_data
- Explanation: If the value is not valid JSON, the node silently treats it as an empty stack. An entry is also skipped when it is disabled, lacks a
loraname, or has effective model and CLIP strengths of zero. - Solution: Use the frontend controls or check that
loras_datais a valid JSON array with the intended names, enabled settings, and strengths.
CLIP effect is missing
- Explanation: Without a
clipconnection, the node forces every entry's CLIP strength to zero. This is expected behavior even when a CLIP strength is stored in the stack. - Solution: Connect the appropriate
CLIPoutput to the optionalclipinput if you want LoRAs applied to text conditioning.
ValueError: could not convert string to float: ...
- Explanation: A strength value that cannot be converted to a number can stop execution while the node processes that entry.
- Solution: Correct the entry's
strengthModel,strengthClip, or fallbackstrengthvalue to a numeric value, preferably through the frontend controls.
AttributeError: '...' object has no attribute 'get'
- Explanation: The JSON was parsed, but an item in the stack was not an object with fields such as
loraandenabled. The node tries to read those fields from every entry it encounters. - Solution: Repair the JSON so the top-level value is an array of entry objects, or recreate the affected row through the frontend.
LoRA file loading or application fails
- Explanation: After finding a file, the node passes it to ComfyUI's file loader and LoRA application functions. Those operations may raise their own errors; the supplied class does not define a specific message or recovery behavior for them.
- Solution: Read the full ComfyUI error to identify the affected file, then check that the LoRA file can be loaded and is suitable for the model and workflow you are using.
