What operations work best for different types of photos?
Need tips for using the right algorithm for your photos?
In order to produce a high-quality image, it is crucial to make sure it wasn’t previously upscaled using any tool
💡 If your image was already upscaled or scanned from the physical original, sometimes decreasing resolution before processing can yield much better results.
The core basics about Let's Enhance algorithms
Let's Enhance algorithms help to improve your photos in multiple ways: apply JPEG noise reduction system (removing pixelation in your compressed images), bring out shapes, make your images crisp & sharp, replenish missing details, you name it. However, to get the best results, it is worth to know what operations to choose for relevant photo types: taken by phone, digital camera, small compressed & pixelated, for printing, photos of people, architecture, cartoons, etc.
Choose the right algorithm that is most suitable for your use case
Within the OPERATIONS tab, you can adjust the Upscale type, apply Light AI (HDR effect), enhance Tone and Color. These can be applied accordingly within the menu panel on the right:
1. Upscale type: Photo
Upscaling type Photo algorithm is trained for super advanced restoration of low-quality images taken by your phone and digital camera of:
💡 Using the ‘Photo’ algorithm is not recommended for good and high-quality images since the upscale type is trained for aggressively restoring missing details and intensively redrawing them in low-quality images. That’s why some artifacts can be found in good/high-quality images because of the over-intensive restoration.
Within this upscaling type, you can apply an advanced upsampling feature Enhance face details:
💡 When combined, these two algorithms perform an outstanding job at enhancing photos with faces, removing artifacts, redrawing missing elements, and adjusting the crispness of every line to produce sharp and clear images. The 'Enhance face details' is a separate AI algorithm trained on millions of datasets containing faces, so this neural network consistently recognizes missing elements and determines what needs to be adjusted.
2. Upscale type: Digital Art
Upscaling type Digital Art works best on:
Digital Art settings - use for illustrations of lower quality
3. Upscale type: Smart Enhance
Upscaling type Smart Enhance works best on:
small & bad quality product photos
food for the web
💡 Smart Enhance specializes in low-quality images (compressed, blurry, pixelated, etc.). It works wonders on low-quality product images and images containing faces, however, this neural network cannot detect images with bodies. It is better to use upscaling types 'Photo' or 'Photo 1.0' for the use case mentioned above.
4. Upscale type: Smart Resize
Smart resize This algorithm is trained to enhance images with all kinds of texts, such as inscriptions on product packages, posters, and maps:
There are also other functions within the OPERATIONS tap you can adjust accordingly to your requirements:
Use the Light AI sidebar to balance out colours and fix lightning in your images.
Choose the Tone Enhance sidebar for higher dynamic range.
Adjust Color Enhance to increase saturation, and provide brightness adjustments
💡 Dependance - when to apply Light AI, Tone Enhance and Color Enhance
👉🏻 To prevent artificial lighting of your processed images after selecting Light AI, it is not recommended to apply Tone and Color enhancers
👉🏻 For Tone Enhance or Color Enhance, intensity levels of 15-30 will be good starting points, providing you hadn't previously applied Light AI. However, there's always a space to adjust their intensity further up to when the demanded level/satisfying outcome is achieved. Experiment!
Learn more: Tutorial #1: How to Upscale a Photo in One Minute with Let's Enhance
Up next: What are Presets and which one should I choose for my photos?
Updated on: 19/09/2022