ComfyUI-Prompt-Manager

ExpressionSelector

Append a saved expression from the Expression category after the input prompt.

Prompt Manager · 1 output

LoraStackCombine

Combine up to four LORA_STACK inputs into one LORA_STACK output.

Prompt Manager · 4 inputs · 0 parameters · 1 output

MultiLoraCombine

Combine an optional MULTI_LORA_STACK with optional LORA_STACK A/B/C/D

Prompt Manager · 5 inputs · 0 parameters · 1 output

MultiLoraSplitter

Split a MULTI_LORA_STACK payload into LORA_STACK outputs A/B/C/D.

Prompt Manager · 1 input · 0 parameters · 4 outputs

MultiLoraStackerLM

Multi-slot LoRA stacker with a visual 4-panel UI (A / B / C / D).

Prompt Manager · 0 inputs · 4 parameters · 3 outputs

PromptExtractor

Extract prompts, LoRA configurations, and model paths from images, videos, and workflow files.

Prompt Manager · 5 outputs

PromptGenerator

Generate or enhance prompts using a local LLM via llama.cpp, Ollama, or a connected CLIP/text encoder.

Prompt Manager · 3 inputs · 8 parameters · 2 outputs

PromptGeneratorKillSwitch

Pass-through any value, stop Prompt Generator's llama.cpp server, and unload an Ollama model from memory.

Prompt Manager · 1 input · 1 parameter · 1 output

PromptGenOptions

Configure LLM model, sampling parameters, and system prompt for the Prompt Generator.

Prompt Manager · 1 output

PromptManager

Simple prompt manager for saving and loading prompts by category. Prompt-only, no LoRA support.

Prompt Manager · 1 output

PromptManagerAdvanced

Full-featured prompt manager with multi-slot LoRA stack support, trigger words, and thumbnail browser.

Prompt Manager · 5 outputs

RecipeBuilder

Recipe Builder (Multi Model). Single builder UI with 4 model slots

4 inputs · 2 parameters · 1 output

RecipeBuilderMultiPrompts

Bundle model_a/model_b/model_c/model_d prompts into a JSON string payload.

Prompt Manager · 4 inputs · 0 parameters · 1 output

RecipeExtractor

- Reads and normalizes workflow metadata into reusable recipe data - Converts raw generation metadata into a structured recipe format that is easier to edit and pass between nodes - Helps preserve key settings (for example model family, sampler context, and recipe-specific parameters) when moving from ad-hoc runs to reusable setups - Useful when you want to recreate a look/style from existing outputs and then iterate from a stable baseline

RecipeManager

- Save, browse, and reuse recipe entries - Provides persistent recipe storage so successful setups can be cataloged and recalled quickly - Works with Builder to quickly review and update saved recipe data - Encourages a library-style workflow: capture working recipes, name/tag them, and reapply them across projects - Great for maintaining repeatable production presets while still allowing safe versioned edits over time

RecipeModelLoader

- Loads model/checkpoint assets based on recipe data - Resolves and applies model selections encoded in recipes so execution follows the intended model path automatically - Supports model slot selection and loader-specific behavior - Makes recipe playback more reliable across sessions by tying loader behavior to saved recipe intent instead of manual recollection - Useful for multi-model pipelines where model choice is part of the recipe contract, not an afterthought

RecipeModelPicker

Builder-style Type/Model/VAE/CLIP selector that outputs model_data for Recipe Relay.

Prompt Manager · 1 output

RecipeRelay

- Relays recipe data and compatible passthrough context across nodes - Acts as a clean handoff point for recipe payloads so larger graphs stay modular and easier to reason about - Preserves compatible context as data moves through the pipeline, reducing brittle copy/paste style wiring patterns - Helpful when splitting complex graphs into logical stages (extract/build/render/manage) without losing recipe continuity

RecipeRenderer

- Executes recipe-driven generation for fast reruns after setup - Interprets recipe data and applies it to generation execution so you can reproduce prior results or batch variants with minimal manual changes - Reduces setup overhead for repeated jobs by moving configuration logic into reusable recipe data - Especially useful when you want to test multiple prompt/input variations while keeping the same core generation recipe intact