Nodes/Eric_Krea2/Eric Krea2 Multi-LoRA Stack
ComfyUI Node

Eric Krea2 Multi-LoRA Stack

The ten-slot LoRA stack where every slot knows which stage it's on

By EricRollei·Created 4 months ago·Updated 2 days ago· 23
Eric Krea2 Multi-LoRA Stack
  • pipeline
  • suggestions
  • pipeline
  • settings
  • trigger_words
  • prompt
◄lora_presetcustom►
◄ephemeraltrue►
◄on_1true►
◄lora_1none►
◄strength_1s11.00►
◄strength_1s21.00►
◄strength_1s31.00►
◄on_2true►
◄lora_2none►
◄strength_2s11.00►
◄strength_2s21.00►
◄strength_2s31.00►
◄on_3true►
◄lora_3none►
◄strength_3s11.00►
◄strength_3s21.00►
◄strength_3s31.00►
◄on_4true►
◄lora_4none►
◄strength_4s11.00►
◄strength_4s21.00►
◄strength_4s31.00►
◄on_5true►
◄lora_5none►
◄strength_5s11.00►
◄strength_5s21.00►
◄strength_5s31.00►
◄on_6true►
◄lora_6none►
◄strength_6s11.00►
◄strength_6s21.00►
◄strength_6s31.00►
◄on_7true►
◄lora_7none►
◄strength_7s11.00►
◄strength_7s21.00►
◄strength_7s31.00►
◄on_8true►
◄lora_8none►
◄strength_8s11.00►
◄strength_8s21.00►
◄strength_8s31.00►
◄on_9true►
◄lora_9none►
◄strength_9s11.00►
◄strength_9s21.00►
◄strength_9s31.00►
◄on_10true►
◄lora_10none►
◄strength_10s11.00►
◄strength_10s21.00►
◄strength_10s31.00►
◄stack_enabledtrue►
◄fetch_triggerstrue►
◄add_triggersnone►
◄force_refetchfalse►
◄prompt►
◄word_options►
◄trigger_joincomma►

This is the pack's answer to "I need three LoRAs and they don't all matter at the same time." Eric Krea2 Multi-LoRA Stack is a single growable node holding up to ten LoRA slots, and each slot carries its own three weights - one for Stage 1, Stage 2, Stage 3 of a multi-stage run. Apply LoRA is the one-LoRA version of the same idea; this is the version for when a style LoRA should own composition, a face LoRA should carry the refine pass, and a detail LoRA should only touch the final stage.

The mechanics are simple: pick a file per slot from models/loras/Krea2/, and the node fuses the enabled adapters onto the pipeline in slot order. Unused rows auto-hide until you need them, which keeps the panel from being a wall of widgets. Each slot gets:

  • on_N - enable/disable without clearing the slot (distinct from none, which is an empty slot).
  • lora_N - the file dropdown.
  • strength_Ns1 / s2 / s3 - the per-stage weights. Set all three equal for a uniform strength; spread them for stage-scoped control.

Ten slots is a lot. If you're stacking more than four or five, the honest question is whether your LoRAs are doing real work or fighting each other - but the per-stage spread is precisely what makes layered stacks behave, since each adapter can own the stage it's actually good for instead of applying flat everywhere.

Shared controls worth knowing:

  • ephemeral (default ON) - load fresh, unload after the run. Off leaves everything resident and forces manual unloading; the tooltip's "if ANY LoRA in the stack is ephemeral, the whole stack is cleared after the run" is the reminder that this is all-or-nothing.
  • stack_enabled - master switch for the whole node; pass the pipeline through untouched without clearing your slots.
  • lora_preset - ★ Save preset dropdown, storing slots, strengths, and toggles. The shipped samples (Split_Distil, turbo-distill-128-split) hint at the split LoRA tricks people run on Turbo - the community's Raw + Turbo LoRA at 0.6 path is a one-preset recipe.
  • fetch_triggers / add_triggers / force_refetch - trigger words for every enabled slot, combined and de-duplicated in slot order into the trigger_words output, optionally merged into the prompt passthrough. Only the first lookup per file touches the network (cached in data/lora_triggers.json).
  • prompt - passthrough so merged triggers reach your prompt output.

Outputs mirror Apply LoRA: pipeline, settings, trigger_words, prompt. Wire pipeline into Generate or Multi-Stage Ultra; fold settings into Merge Settings for the reproducibility chain.

Where people get burned: per-stage weights are easy to forget once you have ten slots of them - a slot left at 1.0 across all three stages will apply its full strength at every resolution, and a heavy style LoRA at Stage 3 is a common source of that "looks great at S1, ruined at 4k" complaint. And with ephemeral off, orphaned weights silently follow you between workflows - the Unload node exists for exactly that.

How to install

Part of EricRollei/Krea2_ComfyUI_Advanced - ComfyUI Manager → search "Krea2_ComfyUI_Advanced", or:

cd ComfyUI/custom_nodes
git clone https://github.com/EricRollei/Krea2_ComfyUI_Advanced

Restart. Pack-wide requirements: git-installed diffusers with Krea2Pipeline, transformers>=4.57.0, Krea 2 weights, LoRAs in models/loras/Krea2/.

Common issues

  • Slots disappearing - unused rows auto-hide; pick a file to activate a slot, and a new empty one appears below.
  • Weights bleeding between runs - ephemeral off. Flip it on or unload manually.
  • Trigger words never appear - fetch_triggers must be on, and the node only queries Civitai for files without safetensors metadata; force_refetch is there for stale lookups.
  • Stacked LoRAs muddy each other - spread them per-stage instead of stacking all at 1.0 everywhere.
CategoryEric/Krea2

Inputs (61)

NameTypeDefaultDescription
pipelineKREA2_PIPELINE—
lora_presetoptCOMBOcustomLoad a saved LoRA-stack recipe (slots / strengths / toggles) into the panel for this run. 'custom' = use the panel as-is. Use the ★ Save Preset button to add one; the list refreshes on graph reload.
ephemeraloptBOOLEANtrueON (recommended): load these LoRAs fresh each run and unload them after generation (no stale/orphaned weights). OFF: leave them resident (faster reuse, but Unload manually to free VRAM / before switching workflows). Applies to every slot in this node.
on_1optBOOLEANtrueEnable/disable LoRA slot 1 without clearing it. OFF keeps the file and strength but skips loading it (distinct from 'none' = empty slot).
lora_1optCOMBOnoneLoRA slot 1 (from ComfyUI/models/loras/Krea2/). Pick a file to activate it - a new empty slot appears below. 'none' = unused/off.
strength_1s1optFLOAT1.00-10–50000Slot 1 weight during Stage-1 (composition). Set all three equal for a uniform strength.
strength_1s2optFLOAT1.00-10–50000Slot 1 weight during Stage-2 (refine).
strength_1s3optFLOAT1.00-10–50000Slot 1 weight during Stage-3 (final detail).
on_2optBOOLEANtrueEnable/disable LoRA slot 2 without clearing it. OFF keeps the file and strength but skips loading it (distinct from 'none' = empty slot).
lora_2optCOMBOnoneLoRA slot 2. 'none' = unused/off.
strength_2s1optFLOAT1.00-10–50000Slot 2 weight during Stage-1 (composition). Set all three equal for a uniform strength.
strength_2s2optFLOAT1.00-10–50000Slot 2 weight during Stage-2 (refine).
strength_2s3optFLOAT1.00-10–50000Slot 2 weight during Stage-3 (final detail).
on_3optBOOLEANtrueEnable/disable LoRA slot 3 without clearing it. OFF keeps the file and strength but skips loading it (distinct from 'none' = empty slot).
lora_3optCOMBOnoneLoRA slot 3. 'none' = unused/off.
strength_3s1optFLOAT1.00-10–50000Slot 3 weight during Stage-1 (composition). Set all three equal for a uniform strength.
strength_3s2optFLOAT1.00-10–50000Slot 3 weight during Stage-2 (refine).
strength_3s3optFLOAT1.00-10–50000Slot 3 weight during Stage-3 (final detail).
on_4optBOOLEANtrueEnable/disable LoRA slot 4 without clearing it. OFF keeps the file and strength but skips loading it (distinct from 'none' = empty slot).
lora_4optCOMBOnoneLoRA slot 4. 'none' = unused/off.
strength_4s1optFLOAT1.00-10–50000Slot 4 weight during Stage-1 (composition). Set all three equal for a uniform strength.
strength_4s2optFLOAT1.00-10–50000Slot 4 weight during Stage-2 (refine).
strength_4s3optFLOAT1.00-10–50000Slot 4 weight during Stage-3 (final detail).
on_5optBOOLEANtrueEnable/disable LoRA slot 5 without clearing it. OFF keeps the file and strength but skips loading it (distinct from 'none' = empty slot).
lora_5optCOMBOnoneLoRA slot 5. 'none' = unused/off.
strength_5s1optFLOAT1.00-10–50000Slot 5 weight during Stage-1 (composition). Set all three equal for a uniform strength.
strength_5s2optFLOAT1.00-10–50000Slot 5 weight during Stage-2 (refine).
strength_5s3optFLOAT1.00-10–50000Slot 5 weight during Stage-3 (final detail).
on_6optBOOLEANtrueEnable/disable LoRA slot 6 without clearing it. OFF keeps the file and strength but skips loading it (distinct from 'none' = empty slot).
lora_6optCOMBOnoneLoRA slot 6. 'none' = unused/off.
strength_6s1optFLOAT1.00-10–50000Slot 6 weight during Stage-1 (composition). Set all three equal for a uniform strength.
strength_6s2optFLOAT1.00-10–50000Slot 6 weight during Stage-2 (refine).
strength_6s3optFLOAT1.00-10–50000Slot 6 weight during Stage-3 (final detail).
on_7optBOOLEANtrueEnable/disable LoRA slot 7 without clearing it. OFF keeps the file and strength but skips loading it (distinct from 'none' = empty slot).
lora_7optCOMBOnoneLoRA slot 7. 'none' = unused/off.
strength_7s1optFLOAT1.00-10–50000Slot 7 weight during Stage-1 (composition). Set all three equal for a uniform strength.
strength_7s2optFLOAT1.00-10–50000Slot 7 weight during Stage-2 (refine).
strength_7s3optFLOAT1.00-10–50000Slot 7 weight during Stage-3 (final detail).
on_8optBOOLEANtrueEnable/disable LoRA slot 8 without clearing it. OFF keeps the file and strength but skips loading it (distinct from 'none' = empty slot).
lora_8optCOMBOnoneLoRA slot 8. 'none' = unused/off.
strength_8s1optFLOAT1.00-10–50000Slot 8 weight during Stage-1 (composition). Set all three equal for a uniform strength.
strength_8s2optFLOAT1.00-10–50000Slot 8 weight during Stage-2 (refine).
strength_8s3optFLOAT1.00-10–50000Slot 8 weight during Stage-3 (final detail).
on_9optBOOLEANtrueEnable/disable LoRA slot 9 without clearing it. OFF keeps the file and strength but skips loading it (distinct from 'none' = empty slot).
lora_9optCOMBOnoneLoRA slot 9. 'none' = unused/off.
strength_9s1optFLOAT1.00-10–50000Slot 9 weight during Stage-1 (composition). Set all three equal for a uniform strength.
strength_9s2optFLOAT1.00-10–50000Slot 9 weight during Stage-2 (refine).
strength_9s3optFLOAT1.00-10–50000Slot 9 weight during Stage-3 (final detail).
on_10optBOOLEANtrueEnable/disable LoRA slot 10 without clearing it. OFF keeps the file and strength but skips loading it (distinct from 'none' = empty slot).
lora_10optCOMBOnoneLoRA slot 10. 'none' = unused/off.
strength_10s1optFLOAT1.00-10–50000Slot 10 weight during Stage-1 (composition). Set all three equal for a uniform strength.
strength_10s2optFLOAT1.00-10–50000Slot 10 weight during Stage-2 (refine).
strength_10s3optFLOAT1.00-10–50000Slot 10 weight during Stage-3 (final detail).
stack_enabledoptBOOLEANtrueMaster switch for the WHOLE stack. OFF = pass the pipeline through untouched (no slot is queued, no triggers emitted) without clearing any slot settings.
fetch_triggersoptBOOLEANtrueLook up trigger words for every ENABLED slot (safetensors metadata / Civitai, cached locally in Eric_Krea2/data/lora_triggers.json - only the first lookup per file touches the network). Fills the trigger_words output (combined, de-duplicated, slot order).
add_triggersoptCOMBOnoneWords added to the prompt for each LoRA, from the LoRA Catalog: none = nothing. trigger = the LoRA's main trigger from its card (your priority trigger first; can be a few words, e.g. 'retro artstyle illustration'). Enough on the base model, often too weak on fine-tunes. full = the whole recipe: official trained words + trigger + style phrase (no card recipe: all its card triggers). Strongest; uses more of the prompt. saved pick = the words chosen for the loaded checkpoint - yours (card editor / Sweeps page), else a trigger test's - else trigger + style phrase. A pick of 'no words' adds nothing on purpose. With two or more LoRAs, full and saved pick are merged into one tidy list (repeats dropped, length shared by weight, clashes reported). Words already in the prompt are not added twice; all are put in front of the prompt. Applies to every enabled slot; the words go into the 'prompt' output (connect your prompt to the prompt input).
force_refetchoptBOOLEANfalseBypass the local trigger cache and re-query Civitai for the enabled slots this run. Leave OFF normally.
promptoptSTRINGOptional prompt passthrough. Connect your prompt here and take the 'prompt' output onward; with add_triggers on, the enabled slots' trigger words are merged in.
suggestionsoptLORA_SUGGESTIONSOptional: LoRAs picked by the LoRA Suggester (video_prompter). They are added after this panel's own slots with their per-stage weights; a file already in the stack is not added twice. Their trigger words join the trigger merge.
word_optionsoptSTRINGPick one of a LoRA's word options (LoRA Catalog card > Prompt words > Word options) per slot, one line each: '3: teal orange' (a name, its number, or the start of its name). That slot's words become the option's words, in every add_triggers mode except none; the trigger_words output carries them too. Use the 'Pick words' button to choose from a list. '3: random' = a different option every run. An amber dot left of a slot = its LoRA has word options and none is chosen (click it to pick). Empty = every slot uses its usual words.
trigger_joinoptCOMBOcommaHow the trigger words join the prompt (when add_triggers is on). comma: 'trigA, trigB, your prompt' (the behaviour before 2026-10-29). period: each LoRA's words become their own sentence - 'trigA. trigB. Your prompt' - the way many Civitai prompts start; it may stop a trigger being read as text to draw (being tested with the Sweep Plan trigger_joins axis).

Outputs (4)

NameTypeDescription
pipelineKREA2_PIPELINE—
settingsSTRING—
trigger_wordsSTRING—
promptSTRING—