Cloudinary’s latest enhancement to its upscale feature significantly advances face detection and improvement capabilities. When using the e_upscale effect, the face upscale feature automatically detects faces and applies additional logic to these areas, ensuring faces upscale with remarkable precision. This advancement holds excellent potential for user-generated content (UGC) and situations where enhancing face-cropped images is crucial.
Cloudinary’s e_upscale feature, launched last year for super-resolution imaging, has transformed the usability of overly cropped or otherwise small images. You can see its initial effectiveness in upscaling landscape or architectural images by viewing examples in the documentation. However, the feature faced a challenge in handling faces. Applying the same logic to faces often resulted in an unnatural, smoothed-over look, which tends to be inauthentic and unappealing to the eye. Given how our brains are wired, faces are the most critical part of an image when present. Ensuring faces appear correctly is critical, as studies show.
Cloudinary’s face upscale feature now enters the scene, automatically detecting human faces when using `e_upscale` and applying specialized generative AI enhancements. This approach allows the rest of the image to benefit from the standard upscale algorithm while treating faces with a finer touch through generative AI. The result? A perfectly upscaled image where faces are clear, natural, and accurate.
Studies show that when people are pictured, faces are the most crucial aspect of an image. Compromising on image quality isn’t an option for brands that rely on images of people to sell products and tell stories. Studies also show that low-quality photos result in a “no purchase” decision by shoppers 30% of the time (source). Now, images with human faces that need restoration or upscaling, like older images or low-quality UGC, can be corrected with extreme attention to the fine details. Cloudinary customers are free to reuse or upscale images confidently without compromising image quality.
- UGC. Companies employing UGC often cannot typically control the quality of original images. The Generative Face Upscale feature ensures any faces in small original images upscales beautifully and consistently.
- Face-cropped images. When using features like g_auto:faces for face cropping, the resultant images might zoom in significantly, resulting in smaller pictures. This feature allows for enlarging these images to a usable size while ensuring the faces look great.
- Expanding small images for reuse. When the only available version of an image is quite small,
e_upscalenow offers the assurance that faces will appear crisp when enlarged.
Face Upscale is already live and ready for use. It can be accessed by using e_upscale as you always have. If the image contains faces, it will auto-detect them and apply the generative upscale to them. The usage is as simple as before, just add e_upscale to your image URLs.
Here’s an example:
Please note that some conditions apply to images that use e_upscale. These usage details have been in place since the initial launch of e_upscale and can be found in the detailed documentation, along with full usage explanations.
Slide to see the upscale effect. The image on right has been upscaled with our generative AI.
Sample upscaled without AI:
Sample upscaled using Generative Face Upscale:
Cloudinary’s face upscale feature marks a significant leap in image superresolution with generative AI. Enhancing faces within images opens up new possibilities for quality and an improved aesthetic in various applications. Regardless of industry, the ability to upscale faces accurately ensures that every image tells a captivating story that resonates with every viewer.
