Doss Multi-LoRA Loader:
DossMultiLoraLoader applies an ordered stack of LoRAs in one place, making it easier to combine several visual influences without setting up a separate loader for each one. You can enable or disable individual entries and adjust their model and CLIP strengths separately. Enabled entries are applied in their listed order, so changing that order can change the final combination. Entries that are disabled or have both strengths set to zero are skipped.
Doss Multi-LoRA Loader Input Parameters:
model
A required MODEL connection supplying the model to which the LoRA stack is applied. Each active entry updates the result passed to the next entry, so the model you connect here is the starting point for the entire stack. There is no numeric minimum, maximum, or default for this connection.
clip
A required CLIP connection supplying the CLIP component to which the LoRA stack is applied. It is passed through the same ordered sequence as the model, with each entry using its own CLIP strength. There is no numeric minimum, maximum, or default for this connection.
lora_stack_json
A required STRING field that stores the editable LoRA stack; the node indicates that you can edit it with controls inside the node. Its default is DEFAULT_LORA_STACK, whose actual JSON contents are not provided here. The stack must be a JSON object with a supported version and an entries list. Each entry needs a LoRA filename in name and can include enabled, strength_model, and strength_clip. Entries are applied in list order. If an entry omits either strength, that strength falls back to its strength value, or to 1.0 if strength is also absent. Both strengths must be finite numbers from -100 to 100; the permitted number of entries is capped by MAX_LORA_ENTRIES, whose numeric value is not provided. The field itself has no stated character-length limit.
Doss Multi-LoRA Loader Output Parameters:
model
The resulting MODEL after all active LoRA entries have been applied in order. Connect it to downstream nodes that need the LoRA-adjusted model. If every entry is disabled or has both strengths set to zero, this is the original model connection.
clip
The resulting CLIP after the same active entries have been applied in order, using their CLIP strengths. Connect it to downstream nodes that need the corresponding LoRA-adjusted CLIP component. If no entry is applied, this is the original CLIP connection.
Doss Multi-LoRA Loader Usage Tips:
- Arrange entries in the order you want them applied. The node passes each entry’s result to the next, rather than applying every LoRA independently to the original connections.
- Adjust
strength_modelandstrength_clipseparately when you want different weights for the model and CLIP. Setting both to0skips loading and applying that entry. - Disable entries you are not currently using instead of removing them from your stack. Disabled entries are skipped during execution.
- Check that each
namematches a LoRA filename available through your ComfyUI LoRA paths. A missing file will prevent that active entry from loading. - The node caches loaded LoRA files and reloads a cached file when its modification time or size changes. You do not need to rebuild the stack just to use an unchanged file again.
Doss Multi-LoRA Loader Common Errors and Solutions:
LoRA stack is not valid JSON.
- Explanation: The stack text cannot be parsed as JSON.
- Solution: Check for missing quotes, commas, or braces. If you edit the text directly, validate it as JSON before running the node.
LoRA stack must be a JSON object.
- Explanation: The stack parses successfully, but its top level is not an object.
- Solution: Use an object containing
versionandentries, rather than a top-level list or text value.
LoRA stack version must be 1 or the current schema version.
- Explanation: The
versionvalue is neither1nor the supported current schema version. The current version’s numeric value is not shown in the provided source. - Solution: Use the node’s stack controls to produce a supported stack, or set
versionto1if you are preparing compatible JSON manually.
LoRA stack entries must be a list.
- Explanation: The
entriesfield is missing or is not a JSON array. - Solution: Provide
entriesas an array, such as[]for an empty stack.
LoRA stack cannot exceed the maximum number of entries.
- Explanation: The stack contains more than
MAX_LORA_ENTRIESentries; the numeric limit is not shown in the provided source. - Solution: Remove entries until the stack is within the node’s limit.
LoRA entry N must be an object.
- Explanation: Entry
Nis a value such as text or a number instead of a JSON object. - Solution: Replace that entry with an object containing at least a
namefield.
LoRA entry N is missing a filename.
- Explanation: Entry
Nhas noname, or its name is empty after surrounding spaces are removed. - Solution: Choose a LoRA file for that entry and ensure its
nameis not blank.
LoRA entry N has an invalid MODEL or CLIP weight.
- Explanation: A strength value for entry
Ncannot be converted to a number. - Solution: Enter numeric values for
strength_modelandstrength_clip, or correct the legacystrengthvalue if the entry relies on it.
LoRA entry N MODEL or CLIP weight must be finite and between -100 and 100.
- Explanation: The indicated weight is outside the allowed range or is not a finite number.
- Solution: Set that weight to a finite number from
-100through100.
LoRA file cannot be found or loaded.
- Explanation: An active entry’s filename cannot be resolved through the configured LoRA paths, or its file fails to load. The exact error text depends on the underlying file-loading operation.
- Solution: Verify the filename and file location, then check that the LoRA file is valid and readable.
