We're excited to announce the release of Autoenhance.ai V3.3, a significant upgrade that builds on our commitment to providing professional quality real estate image enhancements. Our latest version pushes the boundaries of what's possible with artificial intelligence, delivering even better enhancements. Over the past half year we have been talking extensively with a lot of customers to hear about their feedback. On top of that, professional real estate image editors have reviewed each step of our enhancement pipeline so we knew exactly where to improve. Based on the responses of our customers and professional editors, we knew we had some work to do.
Watch a presentation from Jamie, our CEO, about our focus and what we’re working towards at Autoenhance.ai.
Let’s dive into our latest updates. We’re happy to share V3.3 is now able to:
Create more realistic and vibrant colors
A very challenging task while automating real estate image enhancements, is to prevent that the enhancement looks like a general filter that’s been applied to the entire image. Experienced image editors know that walls need to be edited differently than floors for example, which results in a more realistic enhancement. By using advanced deep learning methods, our AI is now able to improve colors on different parts of the image.
Improve local contrast, textures and exposure
V3.3 is able to identify subtle variations in contrast and textures to enhance the overall visual appeal. By employing advanced neural networks, V3.3 excels at preserving fine details while intelligently boosting local contrast. This results in improved textures, for example in wooden floors or furniture, which contributes to a heightened sense of realism. We also improved the way the images are exposed, resulting in less overexposure and better overall balance.
Optimize perspective correction
With V3.3 we completely rebuilt the way perspectives are corrected in an image. Autoenhance.ai is now able to understand the structures in an image and has learned which elements need to be straight. It also knows when an image should only correct verticals, horizontals or nothing at all. Read more about this challenge here.