Compress by quality or target size
With target size set to zero, the tool encodes once at the selected quality. With a target in kilobytes, it repeatedly encodes the same pixels and searches for the highest tested quality at or below both the selected quality ceiling and target. The result reports whether the target was reached. Exact byte size can vary by browser encoder.
Resize before discarding more detail
Pixel dimensions often matter more than the quality slider. A camera photo several thousand pixels wide may be unnecessarily large for a website thumbnail, form upload, or email. Setting a maximum dimension scales width and height proportionally without enlarging a smaller image. Try an appropriate display size before forcing quality very low.
JPEG and WebP behave differently
Both outputs are lossy. JPEG is broadly compatible and works well for photographs, but it cannot store transparency; transparent pixels are placed on white. WebP often produces a smaller photographic file at comparable visual quality and can preserve transparency, though results depend on image content and browser support. Logos, screenshots, and text-heavy graphics may show artifacts sooner than photographs.
Why a target may be impossible
Quality has a practical floor. If the lowest tested quality is still larger than the requested target, the tool returns that smallest candidate and tells you the target was missed. Reduce the pixel dimensions or choose a less aggressive target. A target is a byte constraint, not a promise of acceptable visual quality.
Privacy and metadata
Decoding and encoding use browser canvas APIs, and Gypes does not intentionally upload either image. Canvas output normally omits EXIF fields such as camera model, capture time, orientation, and GPS location. Browser color conversion may also remove profiles or change appearance, so compare the saved file and keep the original.
Compression is destructive. Inspect fine text, faces, gradients, and edges at full size before using the output. This tool does not preserve animation, EXIF metadata, color profiles, or archival fidelity, and it cannot guarantee a smaller result than every source file.