Using LoRAs in ComfyUI Workflows
ComfyUI
Demo summary
The video demonstrates how to insert a 'Load LoRA' node into a workflow to influence a base model's style or character consistency.
Step-by-step
- Add a Load Checkpoint node to select your main base model
- Insert a Load LoRA node between the Load Checkpoint and the K Sampler
- Connect the Model output from the Load Checkpoint to the Model input of the Load LoRA node
- Connect the Clip output from the Load Checkpoint to the Clip input of the Load LoRA node
- Connect the Model and Clip outputs from the Load LoRA node to your Prompt and K Sampler nodes
Options
- Use a Triple Clip Loader (Clip L, Clip G, and T5XXL) instead of the default checkpoint clip
- Load a separate Clip node if it is not embedded in your safe tensor file
Watch out for
- The Load LoRA node must be placed before the K Sampler to influence the generation
- If using an advanced Triple Clip Loader, connect it before the LoRA node and bypass the checkpoint's internal clip
Tips
- Use a Triple Clip Loader to provide the K Sampler with more 'meat' and variable vectors for better prompt adherence
- Note that most safe tensor files have a clip already loaded inside, so a separate loader is often unnecessary for basic workflows
All demos from “ComfyUI Explained: How AI Image Generation Actually Works (Step-by-Step)”
- 13:575:43Configuring the K Sampler in ComfyUIA walkthrough of the K Sampler node settings, including seed generation, step count, CFG (classifier free guidance), and sampler algorithms like Euler.ComfyUI· AI Image Generator
- 24:012:38Using LoRAs in ComfyUI WorkflowsCurrentThe video demonstrates how to insert a 'Load LoRA' node into a workflow to influence a base model's style or character consistency.ComfyUI· AI Image Generator
- 28:561:50Image-to-Image Workflow in ComfyUIThe demo shows an image-to-image setup using a 'Load Image' node and a VAE Encode node to use an existing image as a structural guide for the K Sampler.ComfyUI· Image to Image
- Watch “ComfyUI Explained: How AI Image Generation Actually Works (Step-by-Step)” →
AI Image Generator
1:560:39Load a text-to-image workflow and download missing modelsThe user demonstrates how to load a default text-to-image template in ComfyUI and use the 'show missing models' feature to download required AI checkpoints.Kevin Stratvert- 4:110:29Manage generated assets in ComfyUIThe presenter demonstrates how to access the assets panel using the 'A' shortcut to view, save, or delete previously generated images.Kevin Stratvert
15:005:45Generate image with Hermes Agent and ComfyUIThe host uses the Hermes Agent TUI to request a 1024x1024 image of a tree using the Zimage Turbo model, which the agent executes by generating and running a ComfyUI workflow.ComfyUI
6:420:59Configure Gemma 2B Text Encoder in ComfyUIThe creator shows how to load the Gemma 2B text encoder model into the ComfyUI workflow and select it from the node dropdown menu.AI Search
4:270:30Copying and replacing prompts in ComfyUIThe user demonstrates a custom node feature that allows hovering over prompt examples to copy and then replace the current prompt in the workflow with a single click.pixaroma- 13:030:51Managing node colors with PixaRoma nodesThe video demonstrates UI updates for PixaRoma nodes in ComfyUI, including selecting color swatches, copying/pasting colors between nodes, and setting favorite colors.pixaroma
4:150:35Launch ComfyUI on WindowsThe creator demonstrates how to launch the portable version of ComfyUI by running the 'run_nvidia_gpu' batch file and accessing the interface via a web browser.WINBUSH- 10:071:30Browse and load ComfyUI templatesThe creator demonstrates how to navigate the templates menu, filter for local models, and load a Flux 2 image generation workflow.WINBUSH
0:000:33Configuring MSR mode in LTX DirectorThe user demonstrates setting up the Director node in ComfyUI for MSR (Multiple Subject Reference) mode, including bypassing clean latent slices and enabling the MSR LoRA.RedShade- 5:090:19Loading LTX Director V5 workflow templateThe video shows how to search for and load the LTX Director V5 template within ComfyUI and input prompts for character generation.RedShade
4:251:21Iterative rerolling at specific generation stagesThe video shows how to use the 'reroll' button at specific stages of the workflow to change motion or visual quality without re-generating the initial seed samples.Fox•Fur•Essence Films
9:173:04Install ComfyUI using Easy InstallThe instructor demonstrates how to download the ComfyUI Easy Install zip from GitHub, extract it, and run the BAT file to set up a portable local installation.pixaroma