Generate Product Variations From a Reference Image
Use one source image to explore new backgrounds, colorways, and campaign scenes while keeping the product structure recognizable.
Drop image or click to browse
PNG, JPG, or WebP up to 10MB
Create a Variation
Your variation will appear here
Reference-Guided Variation Examples
Generate Variations in 4 Steps
Upload a reference image
Upload one clear main image so the model can preserve shape, framing, and key product details.
Add optional style references
Attach up to three more images when the output needs a specific campaign look, material cue, or product family style.
Describe the change
Write a concise prompt that explains the setting, colorway, material, lighting, or scene direction you want.
Review and continue
Compare the result for edge quality, product accuracy, and scene consistency, then continue into editing or layered handoff.
Frequently Asked Questions
What is image conditioned generation?
It creates a new image from a source image and a prompt, helping you preserve the main subject while changing the surrounding scene, styling, or presentation.
How is it different from text-only generation?
Text-only generation starts from a written prompt. This workflow uses a reference image, so the result can stay closer to the original product, pose, or composition.
Can I keep logos and proportions?
The workflow is designed to preserve recognizable product structure, but final outputs should still be reviewed when logos, labels, or small text need to remain exact.
What inputs are supported?
Clean JPG, PNG, and WebP files work best, especially when the subject is well lit and clearly separated from the background.
Can I add more than one reference image?
Yes. You can add up to three optional style references when you want stronger consistency across a campaign, product family, material, or brand look.
Can I change only the background?
Yes. This workflow is a strong fit for background swaps, seasonal scenes, lifestyle settings, and contextual updates around the same subject.
Is it useful for ecommerce imagery?
Yes. It is useful for product variations, catalog refreshes, seasonal campaigns, marketplace testing, and ad concepts from one source photo.
How many variations should I generate?
Start with a small batch of outputs, compare the strongest directions, then iterate on the best one with a tighter prompt.
What file formats do I get?
Outputs are typically returned as PNG or JPG. You can move the result into Image to Layers if you need a more editable production workflow.
When should I use Image to Layers after generation?
Use Image to Layers when you want to separate foreground elements, refine shadows, isolate objects, or hand the result off for layered editing.
How do I reduce artifacts?
Use clean references, keep prompts specific, avoid asking for too many changes at once, and iterate in small steps so the model stays close to the source composition.
Can teams use this in repeatable workflows?
Yes. Teams can use it for repeatable catalog, campaign, and concept workflows where many products need the same style treatment.