Skip to main content
Krea 3 is the hosted image model from Krea. It is built for instruction-driven editing: you describe what should change and what should stay the same, and the model returns the edited image. Connect no image and it generates from text instead. Krea 3 runs in ComfyUI as a Partner Node, so there are no weights to download and nothing runs on your own GPU. The node sends your prompt and your reference images to Krea’s servers and returns the finished image to your workflow.

​
What Krea 3 is good at

  • Instruction-driven edits: write the change in plain language (“replace the green car with an orange one”) and Krea 3 returns the whole image, keeping what you did not ask to change
  • Text to image from the same node: leave the image inputs empty and the node generates from the prompt alone
  • Up to 8 reference images: connect them in the order you want to address them, then mention them in the prompt as @Image1, @Image2, and so on
  • Two thinking modes: fast is the default and the recommended setting, while high plans complex edits in more detail and takes longer
  • Precise edits: turn on precise_edit to keep everything outside the edited area unchanged
  • Resolution control: 1K or 2K output, with auto aspect ratio following the first reference image
To use the Partner Nodes, you need to ensure that you are logged in properly and using a permitted network environment. Please refer to the Partner Nodes Overview section of the documentation to understand the specific requirements for using the Partner Nodes.
Make sure your ComfyUI is updated.Workflows in this guide can be found in the Workflow Templates. If you can’t find them in the template, your ComfyUI may be outdated.If nodes are missing when loading a workflow, possible reasons:
  1. You are not using the latest ComfyUI version (Nightly version)
  2. Some nodes failed to import at startup

​
Krea 3 text to image workflow

Generate an image from a prompt alone. The Krea 3 Image node has no image connected, so it uses the prompt and the aspect ratio setting. Krea 3 text to image workflow

Run on Comfy Cloud

Open in Comfy Cloud

Download Workflow

Download JSON or search “Krea 3” in Template Library
  1. Write your prompt in the Krea 3 Image node. The shipped example is a single detailed sentence describing the subject, the lighting, and the framing
  2. Set aspect_ratio. Keep it on auto to follow the prompt, or pick one of the ten supported ratios
  3. Pick resolution: the template ships with 1K, and 2K gives a larger canvas at a higher cost
  4. Leave thinking_mode on fast, or switch to high when the prompt describes a complex scene
  5. Click Queue. The image is saved by the SaveImageAdvanced node to ComfyUI/output/

​
Krea 3 image edit workflow

Edit an existing image from an instruction prompt. The reference image is connected to the Krea 3 Image node, and the prompt describes the result you want. Krea 3 image edit workflow

Run on Comfy Cloud

Open in Comfy Cloud

Download Workflow

Download JSON or search “Krea 3” in Template Library
Input materials Upload this file to the LoadImage node, or use your own image:

wide_format_digital_render.png

LoadImage node · wide_format_digital_render.png
wide_format_digital_render.png
  1. Load the image you want to edit in the LoadImage node and connect it to the Krea 3 Image node
  2. Write the edit as an instruction. The shipped example replaces the green car in the scene with an orange one and lists what must stay unchanged: composition, background, building, dust, lighting, and perspective
  3. Turn on precise_edit if everything outside the edited area must be preserved. It needs at least one reference image
  4. Adjust aspect_ratio and resolution for the output canvas
  5. Click Queue and compare the input with the result
More reference images The node accepts up to 8 images. Add them in the order you want to refer to them, then mention each one in the prompt as @Image1, @Image2, and so on.

​
Krea 3 node parameters

The Krea 3 Image node (Krea3ImageNode) exposes the following settings:
  • image_1 … image_8: up to 8 reference images, numbered in connection order. Leave them empty for text to image
  • prompt: the instruction or the scene description. Refer to references as @Image1, @Image2, and so on
  • aspect_ratio: auto, or one of 1:1, 4:5, 3:4, 2:3, 9:16, 5:4, 4:3, 3:2, 16:9, 21:9. auto follows the first reference image, or the prompt when no image is connected
  • resolution: 1K or 2K output
  • thinking_mode: fast is recommended, high plans complex edits in more detail and takes longer
  • precise_edit: keeps everything outside the edited area unchanged. It needs at least one reference image
  • seed: seed for the request. Krea rewrites the prompt before generating, so the same seed does not reproduce the same image

​
Notes and limits

  • Prompt rewriting: Krea rewrites your prompt before it generates, which is why a repeated seed does not repeat an image
  • Reference size: each reference is scaled to the output pixel budget before it is uploaded, so the 2K setting keeps more reference detail than 1K
  • Billing: Krea 3 is a Partner Node, so every run is billed to your Comfy account. See Partner Nodes pricing for rates and Concurrency limits for how many requests you can have in flight
  • Version: the Krea 3 Image node ships with ComfyUI PR #17067, and the two templates are part of the template library from workflow_templates 0.11.84