JPG (also known as JPEG) is an extremely common, lossy¹ raster format for grayscale or color images that does not support transparency. JPG was designed for, and works best with, photographic content. Logos, icons, and other digital artwork with sharp boundaries between regions of constant or near-constant color usually do not look good when encoded as JPG. Ringing, blurriness, color distortions, mosquito noise, and other artifacts are common when encoding such images with JPG.
As a raster format, JPG encodes images as a uniform grid of pixels, each of which can be thought of as a small rectangle (usually a square) of a specified color. Taken together this grid of pixels looks like an image when viewed at its native size, but scaling a raster image to a larger size will result in a pixelated or blurry image.
¹ Lossy means that encoding an image and then decoding it again produces a result that is different from the original.
SVG is a modern vector format based on XML that supports transparency, gradients, and animations. SVG is supported by modern web browsers and is the most common vector format in use today by a wide margin.
Most vector editing software can import and export SVG, even if they typically default to other proprietary formats such as Adobe's AI format. Some of the more advanced features of SVG, such as animations and filter effects have less broad support.
SVG is usually the best choice for logos, icons, diagrams, illustrations, and other similar digital artwork. SVG is the preferred export format of Vectorizer.AI, as it supports all of the features of our vectorization engine.
You can drag and drop your image onto the dashed box above, or click on it to open a file selection dialog.
Once your image is uploaded, the vectorization process will start automatically.
The vectorization process is performed on our high-performance servers, to quickly produce a high quality result.
Once the process is complete you will be shown the result in an interactive viewer capable of zooming and panning, so you can inspect it in detail before downloading.
When you are done reviewing you can click the 'Download' button to fetch your result to your computer.
We offer a wide variety of export options and formats, including SVG, that allow you to tailor the result to your specific needs.
Vectorizing an image is easy for the human eye, but surprisingly hard for the computer. Most software that tries to do it produces poor results, with glaring defects. Shapes can be introduced in the result that should not be there, such as anti-aliasing artifacts, or shapes can be missing that should be there, such as small and/or faint features. Even when the shapes are correct, the curves that define the shapes can be poorly chosen. In some cases, the curves simply don't follow the original image very well. In other cases, there are too many curves, or the curves that are present are poorly placed, don't connect with matching tangents when they should, or are represented using the wrong type of curve (e.g., using a quadratic bezier when an elliptical arc would be better).
Each step in the vectorization process is complex and there are many different algorithms that can be used. Many of our competitors use old and simple algorithms that do not produce good results. Some of them only support 2-color vectorization, which significantly limits their usefulness. The Vectorizer.AI vectorization engine is based on our own proprietary research and uses a combination of deep learning and other techniques to produce the best results. Curves are chosen carefully and optimized to fit the underlying image as closely as possible.
We also identify typical shapes like circles, ellipses, rectangles, stars, and triangles and represent them explicitly as such. This makes the results look better, and makes them easier to edit.
A common simplifying choice made when developing a vectorization algorithm is to only support two colors (e.g., black and white). Products built on top of such algorithms are significantly less useful and versatile than full-color vectorization systems. Other systems support more colors but only by repeatedly running a 2-color algorithm on each color separately.
In contrast, the Vectorizer.AI vectorization engine was built from the ground up to support full-color vectorization, including transparency and partial transparency. The Vector Graph underlying our system seamlessly maintains consistency between adjacent shape boundaries while allowing the system to optimize the result for the best possible quality.
Vectorization comes in two main flavors: reconstructive and inspirational.
Reconstructive vectorization is the process of converting a bitmap image that was once created by rasterizing a vector original, into a vector image that is as close as possible to the original. The goal is to reconstruct the original vector art. It is most useful on logos, icons, and other digital graphics where the original vector art is not available.
Inspirational vectorization converts a photograph, painting, or other other similar raster image into a vector image that is inspired by the original, but does not necessarily attempt to reconstruct it exactly. It is more about capturing some artistic essence or spirit of the original, than to reconstruct a platonic ideal.
Our primary focus is on reconstructive vectorization, but we of course also support inspirational.
Most vector formats support embedding raster images inside of them. Doing so creates a 'fake' vector file since it doesn't change the image's fundamental pixel nature. With such results you still can't do things like scale them to a larger size without loss of quality.
So when converting from JPG to SVG, it is very important to actually vectorize the image. This process involves detecting the shapes in the image, fitting curves to them, and exporting the result as a true vector file. The end result does not contain any pixel data and can be scaled to any size without loss of quality.
At Vectorizer.AI, we only support true vectorization.