How to Use

The Vary Object workflow allows you to identify and modify a specific object in your image while preserving the overall look and feel. However, because this workflow relies on a segmentation model, it works best when the object is clearly distinguishable from the background. If your image contains high complexity or multiple overlapping objects, the model may struggle to isolate the correct element.

For best results, choose images where the object you want to vary is well-defined and easy to segment. If you need to vary an object by selecting a region rather than relying on automatic segmentation, you can achieve this directly in the Canvas instead of using a workflow.

Additionally, this workflow is designed to modify an object rather than completely change it into something else. If you’re looking to transform an object into something entirely different or guide the changes in a very specific way, a different workflow would be more suitable.

1. Upload an Image – Start by uploading an image that contains the object you want to modify.
2. Enter a Prompt – Provide a textual description of the object you want to vary. This helps the system identify what should be changed.
3. Select a Model – Choose the generative model that will be used to process and recreate the object.
4. Adjust Vary Strength – Set the level of variation for the object. A lower value makes minor modifications, while a higher value introduces more drastic changes.
5. Specify a Board (Optional) – You can save outputs to a specific board for better organization.
6. Run the Workflow – Once everything is set up, execute the workflow. The system will identify and modify the object while maintaining the overall look and feel of the image.
7. Review and Refine – Check the results and adjust the settings if necessary for better alignment with your vision.

‍

How It Works

1. Object Detection – The system processes the input image and, using the provided prompt, identifies the object to be modified.
2. Bounding Box Creation – A bounding box is generated around the identified object to isolate it for modification.
3. Segmentation and Masking – The workflow refines the detected object by applying segmentation techniques and creating a precise mask.
4. Variation Processing – The identified object is regenerated based on the selected vary strength, using a diffusion-based generative model.
5. Image Blending – The newly modified object is seamlessly blended back into the original image to maintain coherence and realism.‍
6. Final Processing & Resizing – The final image is processed to match the original dimensions and quality before being saved to the specified board or location.