Krea 2 Turbo — Hires + Detailer Workflow (SAM3) — v1.1 — v3.2
Lama Remover
Runtime profile
- Access
- External workflow
- Status
- Listed
- Version
- krea2TurboHiresDetailer_v32.json
- Source
- Civitai
- Updated
- Oct 16, 2026
What you need
- Model files
- 0.1 GiB known subtotal
Remaining hardware or download-size estimates are not available yet.
Requirements
Models
8- Found
Upscaler · 63.9 MB · PTH · Unknown
- Not resolved
Object Detector · Hugging Face
LoRA · 1.70 GB · SAFETENSORS · fp16 · Hugging Face · Comfy-Org/Krea-2
krea2_turbo_fp8_scaled.safetensors
File unverifiedCheckpoint · Hugging Face · Comfy-Org/Krea-2
- Not resolved
VAE · Hugging Face · Comfy-Org/Qwen-Image_ComfyUI
Text encoder · 4.88 GB · SAFETENSORS · Hugging Face · Comfy-Org/Qwen3-VL
sam_vit_b_01ec64.pth
Not resolvedSegmentation · Unknown
sam3.1_multiplex_fp16.safetensors
Not resolvedVision model · Hugging Face · Comfy-Org/Krea-2
Node packs
6- Found
Node pack · Registry
- Found
Node pack · Registry
Node pack · Registry
comfyui-lama-remover
PossibleNode pack · Registry
erosdiffusion-eulerflowmatchingdiscretescheduler
Not resolvedNode pack · Registry
- Found
Node pack · Registry
AI-generated commentary
AI-generated explanation based on source and configuration details. Suggestions are clearly labeled.
Purpose
This ComfyUI workflow uses Krea 2 Turbo to turn a positive text prompt into an image, then can upscale it and redraw faces or text-selected regions; it can also route an existing image through correction stages and save the result as an image.
Flow
The base generator, upscaler, and Face Detailer are on by default; Hires Fix and the SAM3.1 detail and exclusion groups are off by default.
With SAM3.1 enabled, prompts such as `face:99` or `hand:99` select regions, and a detection preview shows the area to be redrawn.
Inputs and outputs
Edit `Positive Prompt` in `Krea2Turbo`; enabling `Load existing image` routes a loaded image into the correction path, and saved output goes to `ComfyUI/output/Krea2Turbo/`.
The Image Comparer slider lets you compare the detailed result with the raw generation.
Setup
To start, load the workflow, edit the positive prompt, optionally add LoRAs in Power Lora Loader, and queue it.
Dependencies
The listed model files are `4x-UltraSharp.pth`, `bbox/face_yolov8m.pt`, `krea2_identity_edit_v1_2.safetensors`, `krea2_turbo_fp8_scaled.safetensors`, `qwen_image_vae.safetensors`, `qwen3vl_4b_fp8_scaled.safetensors`, `sam_vit_b_01ec64.pth`, and `sam3.1_multiplex_fp16.safetensors`
The listed node packs are `comfyui-freememory`, `comfyui-impact-pack`, `comfyui-impact-subpack`, `comfyui-lama-remover`, `erosdiffusion-eulerflowmatchingdiscretescheduler`, and `rgthree-comfy`.
The SAM3.1 and Identity Edit groups are off by default; using them requires `sam3.1_multiplex_fp16.safetensors` and `krea2_identity_edit_v1_2.safetensors`, respectively.
Configuration
To work on an existing image, enable `Load existing image`; the instructions say base generation still runs unless you set `Krea2Turbo` steps to 1.
Suggestion · not verified
To increase output size, adjust the Upscaler `scale_by` setting; the instructions give `0.375` as an example.
Suggestion · not verified
Cautions
Use Fast Groups Bypasser or Ctrl+B to toggle whole groups; bypassing a single node can leave a required input missing.
Suggestion · not verified
For face correction, choose Face Detailer or SAM3.1; the instructions warn that using both on faces redraws the same area twice.
Suggestion · not verified
Does this need to be edited?
Sign in to send an edit request.
Sources
1 source
Sources
1 sourceSource excerpts
1 excerpt
Source excerpts
1 excerptSource context: 652 downloads · Type Workflows · Base model Krea 2
