Gladas' Workflow for Krea 2 Turbo — v1 — v12b

Author notes from gladasWorkflowForKrea_v12b.json · selected release

# How I Use This Workflow ## LoRA Loaders Each **SampleCustomAdvanced** has a matching **Power Lora Loader**: - The **blue LoRA loader** applies to the **blue SamplerCustomAdvanced**. - The **red LoRA loader** applies to the **red SamplerCustomAdvanced**. ## Initial Resolution Choose your initial resolution on the **SDXL Empty Latent Image** node. You can replace this node with your preferred **Empty Latent** node. Connect its latent output to the `latent_image` input on the **blue SamplerCustomAdvanced**. ## Latent Upscale Set `scale_by` on the **Upscale Latent By** node to **2.00** for a **2× upscale** of both width and height. Adjust this value as needed. ## Sampling Settings I use the two **SamplerCustomAdvanced** nodes to manually split sampling around a latent upscale, approximating how I use the **LC Speed** node. ### Settings for Both SamplerCustomAdvanced | Setting | Value | |---|---| | `sampler_name` | `euler` | | `scheduler` | `simple` | ### Blue SamplerCustomAdvanced — Before Upscaling | Setting | Value | |---|---| | `steps` | `6` | | `denoise` | `1.0` | ### Red SamplerCustomAdvanced — After Upscaling | Setting | Value | |---|---| | `steps` | `4` | | `denoise` | `0.65` | The blue SamplerCustomAdvanced runs the first portion of sampling at the initial resolution. The latent is then upscaled, and the red SamplerCustomAdvanced continues sampling on step **7**.

Author notes from gladasWorkflowForKrea_v12b.json · selected release

## ⚠️ Custom nodes used in workflow --- #### [rgthree-comfy](https://github.com/rgthree/rgthree-comfy) * https://github.com/rgthree/rgthree-comfy #### [ComfyUI-Custom-Scripts](https://github.com/pythongosssss/ComfyUI-Custom-Scripts) * https://github.com/pythongosssss/ComfyUI-Custom-Scripts #### [ComfyUI-KJNodes](https://github.com/kijai/ComfyUI-KJNodes) * https://github.com/kijai/ComfyUI-KJNodes #### [ComfyUI-Impact-Pack](https://github.com/ltdrdata/ComfyUI-Impact-Pack) * https://github.com/ltdrdata/ComfyUI-Impact-Pack #### [ComfyUI-Image-Saver](https://github.com/alexopus/ComfyUI-Image-Saver) * https://github.com/alexopus/ComfyUI-Image-Saver #### [comfyui-gps-supplements](https://github.com/Goshe-nite/comfyui-gps-supplements) * https://github.com/Goshe-nite/comfyui-gps-supplements #### [BKWILDCARDS](https://github.com/bkidderz/BKWILDCARDS) * https://github.com/bkidderz/BKWILDCARDS #### [ComfyMath](https://github.com/evanspearman/ComfyMath) * https://github.com/evanspearman/ComfyMath

Author notes from gladasWorkflowForKrea_v12b.json · selected release

I use the red **Power Lora Loader** to load **Wetness Slider** by **Loraholic** to reduce random wet skin. - Set its strength to a negative value between **-1 and -2**. - [Download Wetness Slider](https://civitai.red/models/2738333/wetness-slider-krea2-zit) - or search for it on CivitAI

Author notes from gladasWorkflowForKrea_v12b.json · selected release

### If you are *bypassing* the *red SamplerCustomAdvanced*, make sure to set the *steps* to *zero*. ### Image Saver will take the combined *steps* value of each *BasicScheduler* node.

Author notes from gladasWorkflowForKrea_v12b.json · selected release

## Model Links **vae** - [seedvr2_ema_vae_fp16.safetensors](https://huggingface.co/Comfy-Org/SeedVR2/resolve/main/vae/seedvr2_ema_vae_fp16.safetensors) **diffusion_models** - [seedvr2_3b_int8_convrot.safetensors](https://huggingface.co/Comfy-Org/SeedVR2/resolve/main/diffusion_models/seedvr2_3b_int8_convrot.safetensors) ## Model Storage Location ``` 📂 ComfyUI/ ├── 📂 models/ │ ├── 📂 vae/ │ │ └── seedvr2_ema_vae_fp16.safetensors │ └── 📂 diffusion_models/ │ └── seedvr2_3b_int8_convrot.safetensors ```

Author notes from gladasWorkflowForKrea_v12b.json · selected release

#### The nodes used for SeedVR2 are taken from the searchable template for the seedVR2_3b found when clicking on the templates button typically located on the left side of your ComfyUI canvas. - These nodes are hidden in a subgraph on the template, but these were the default settings. I haven't messed with the settings, so change them at your own peril. - **The default denoise setting is 1.0** - Changing the **denoise** to lower values does seem to reduce the effect, but other than that IDK.