VPI
The Challenge: Adapting vendor-provided images (VPI) to fit your in-house style is a long, laborious process that can often act as a bottleneck to getting content online quickly.The Solution: A suite of tools that empower production teams to easily modify VPI to meet a wide variety of requirements, all built on automation.
When possible, why not adapt Vendor-Provided Images to your brand style? Replace the background, reframe the crop, match the visual qualities, and relight the shots… and be ready to go to market in hours, not days or weeks.
Using our segmentation models and generative AI, Dreem.ai can automate the process. With perfect matting and image analysis, we can modify and integrate the image to perfectly fit any style. That way, you can get to market quicker and save a host of resources while still delivering the high-quality images you are used to. For us at Dreem.ai, it is important to make this a streamlined process, chaining together a host of narrow AI tools we have built to complete a once-demanding process in seconds or minutes, with minimal human interaction.
There are a host of technical details that go into such a conversion, which is why we use many layers of technology to analyze, break down, enhance, and modify all aspects of the image.
Tools we use when converting VPI to your own brand look and feel:
Upscale, deblur, remove compression artifacts, and sharpen
It is not uncommon to receive Vendor-Provided Images that are not high enough resolution, sharp enough, or have other quality issues. Our Latent Upscaler uses generative AI to upscale and clean up the image, adding the quality back in so the end result looks completely real. Using the same method, we can remove compression artifacts for provided images and ensure they are sharp again.
Background replace
Change the background to match the style of your store and brand. Using advanced image matting, perspective matching, and generative AI, we will output high-fidelity imagery in no time.
Reframe / Recrop / Outpaint
With the ability to match and change the framing and even zoom back out and generate the missing parts of the image, we can make the images look even more like your own. We solve this with machine learning vision algorithms for the framing and cropping, and using our generative AI methods, we can even extend the image when needed.
Re-lighting
If the VPI is shot in soft lighting and your brand uses a harsh light from the side, we will match the lighting to your style! By analyzing and predicting the depth and normal vectors of a photograph, we can pull the image apart, modify it, and put it back together with modified lighting using photographic and 3D rendering-inspired technology.
Other visual qualities
Sharpness, contrast, noise, and shadows… these are all part of the image's feeling and nuanced quality.
We use many of the same tools we develop for QC to pick up on the nuances through deep pixel analysis, generative AI, and AI vision analysis tools. These come together to give you a VPI adaptation toolkit that works at scale!