Lossless vs Re-encoding When Stripping EXIF

The lossless approach treats the file as a structure of segments. In a JPEG, metadata lives in application markers (APP1 for EXIF, APP13 for IPTC, and so on); in a PNG, it lives in ancillary chunks like tEXt and eXIf. A lossless cleaner walks the file, copies across only the segments needed to display the image, and drops the metadata ones. The compressed pixel data is never decoded or re-compressed, so the output is bit-for-bit identical to the original in everything that matters visually.

Concretely, a lossless JPEG cleaner keeps the structural markers — the start of image, the quantisation and Huffman tables, the frame header and the compressed scan — and drops APP1 where EXIF and XMP live, APP13 where Photoshop writes IPTC, and any COM comment. It usually keeps APP2 if an ICC colour profile is there, since that is display information rather than personal information. In a PNG the equivalent is keeping IHDR, PLTE, IDAT and IEND while dropping tEXt, zTXt, iTXt, eXIf and tIME; in a WebP it means removing the EXIF and XMP chunks from the RIFF container and leaving the image chunk alone.

The upside is perfect fidelity and a slightly smaller file. The cost is complexity: the cleaner has to understand each format's internal layout precisely, and unusual or newer container formats can be harder to handle safely. Dedicated desktop tools such as ExifTool specialise in this.

The re-encoding approach ignores the file's internal structure entirely. It decodes the image into raw pixels, draws them onto an off-screen canvas, and asks the browser to encode a brand-new file from those pixels. Because the new file is generated from scratch, it inherently contains no EXIF, IPTC or XMP — there was no metadata block to carry over. This is how MetaCleanse produces its clean copies.

It is worth being precise about what changes along the way. The browser encodes to the format you choose at a quality setting you do not control byte-for-byte, so the chroma subsampling and quantisation tables will differ from the camera's. The canvas works in sRGB, so an embedded ICC profile from a wide-gamut camera is not carried over and colours can shift slightly on a colour-managed display. The Orientation flag vanishes, which is why the pixels must be rotated before encoding. File size may rise or fall depending on the original's quality. And because the browser decodes first, a HEIC input comes out as a JPEG, PNG or WebP rather than as HEIC.

The advantage is robustness and simplicity: it works uniformly across formats the browser can decode, with no need to parse each container by hand, and it runs safely client-side. The trade-off appears with lossy formats. Re-encoding a JPEG re-compresses it once, which is not bit-for-bit identical to the original — though at a high quality setting the difference is invisible to the eye. Lossless formats like PNG re-encode without any quality change at all.

Lossless stripping is what dedicated desktop utilities do. ExifTool's -all= option rewrites a file without its metadata while leaving the compressed image data alone; jpegtran -copy none does the same for JPEG and can also rotate a frame losslessly. Windows' Remove Properties and Personal Information belongs to the same family — convenient, though it clears the fields it knows about rather than guaranteeing an empty file, and it has historically left some blocks in place.

Re-encoding is what most graphical and web tools do, because it reuses the same code path they already have for resizing or converting. ImageMagick's -strip flag, an export from macOS Preview, and a browser canvas all end in a fresh encode. The browser version has one advantage the others lack: the decoding and encoding happen inside the page on your own machine, so a file you are cleaning precisely because it is sensitive never travels anywhere to be cleaned.

For the everyday goal — making a photo safe to share online — re-encoding is the pragmatic winner. It is reliable, needs no installation, runs privately in your browser, and a single high-quality JPEG re-compression is imperceptible in a photo destined for a screen. That combination is why MetaCleanse uses it.

Reach for byte-level, lossless stripping when you must preserve the exact original pixels — archival masters, forensic work, camera raw files, or a professional workflow where re-compression and a lost ICC profile are unacceptable. In those cases a specialised desktop tool is the right instrument; for everything else, a clean re-encode gets the job done in seconds.