VECTOFOX FIELD NOTES
Why is my SVG larger than my PNG?
SVG describes shapes, while PNG and JPEG compress pixels. A scene with many small color changes can need far more vector geometry than a flat logo. The file extension does not guarantee a smaller download.
VectoFox documentation · Updated
Compare actual bytes, not the extension
The measured opaque logo shrinks from 1,195 source bytes to 678 SVG bytes in the balanced run. The fox illustration changes from 46,979 to 19,211 bytes. Those are useful outcomes for these specific fixtures, not a rule that applies to every logo.
The synthetic continuous-tone JPEG grows from 18,122 to 494,492 bytes, with 1,339 visible paths. It contains smooth variation and deterministic noise, not a real photograph. It demonstrates how a compact raster texture can need a large collection of traced regions.
| Sample | Preset | Input bytes | SVG bytes | Visible paths | Est. endpoints | Median ms | RGB error |
|---|---|---|---|---|---|---|---|
| Opaque logo on white | balanced | 1,195 | 678 | 4 | 22 | 17.8 | 0.1 |
| Flat fox illustration | balanced | 46,979 | 19,211 | 40 | 654 | 90 | 4.2 |
| Continuous-tone texture (synthetic) | balanced | 18,122 | 494,492 | 1,339 | 23,504 | 192.9 | 12.97 |
Estimated endpoints count path-segment endpoints, not curve handles, and exclude close commands and clip definitions. RGB error is the average channel difference on white (0–255), not a perceptual ranking. Lower counts and smaller files can hide lost features.
Why JPEG noise can become vector shapes
A raster codec can compress similar pixels and recurring patterns efficiently. The tracer instead finds regions and fits curves to their boundaries. Gradients, grain and compression fringes can split the image into many small regions, each with coordinates and fill information.
Compare the intentionally low-quality JPEG fixture with the original opaque PNG. The balanced SVG for the JPEG is 10,721 bytes from a 1,981-byte input. The version traced from the clean PNG is only 678 bytes. Use the clearest original rather than expecting the vector format to repair the compressed source.


A much smaller result can mean information was discarded
The soft-alpha fixture becomes much smaller after tracing, but its halo is not retained faithfully. The outer probe changes from alpha 52 to transparent. The fixed threshold converts partial transparency to opaque regions or removes it, so this small SVG is a simplified design rather than an equivalent compressed copy.
Similarly, reducing a palette can merge shades that matter to the artwork. Review the trace before celebrating a size reduction. File bytes are objective, but they are not a perceptual quality score.

Reduce the right source of complexity
For flat artwork, try fewer colors, less path detail and a source without compression speckles. Inspect the edge and any small text after changing the settings. An intentional logo corner should survive even if the export could be a few bytes smaller without it.
For a photo or continuous-tone image, keeping the original JPEG or WebP may be more appropriate. Vectorization creates a stylized interpretation, not a replacement for photographic detail. Use SVG when editable shapes and scaling are the requirement, not merely because the filename looks more professional.
Keep transfer size separate from file size
The measurements here count the UTF-8 bytes in the uncompressed downloaded SVG. A website may send text assets with HTTP compression, so network transfer size can be different. A browser cache, response headers and server settings also affect repeated visits.
SVG optimization tools can remove redundant markup or reduce coordinate precision, but they cannot guarantee that a complex trace becomes as small as a raster image. Always render a copy after optimization, especially when clip paths preserve transparency. VectoFox currently traces raster images; it is not a general optimizer for uploaded SVG documents.