Why conversion changes more than metadata
The browser decodes the visible image into a canvas and encodes a new file. Lossy output can change fine pixel values even when dimensions stay the same.
When to clean instead
If you want to keep an existing JPEG or supported container payload, use the metadata cleaner. If you need a new format, size or compression level, use conversion and retain the original separately.
Understand the decode-and-encode boundary
Browser conversion reads the visible source into a pixel surface and creates a new file with an encoder. Container blocks around the source pixels are not automatically carried through that surface.
That is why EXIF camera values, GPS and ordinary XMP were absent from the tested outputs. The result is a new derivative, not the original file with a different label.
- Pixels decoded
- Optional resize applied
- New format encoded
- Output signature verified
- Output metadata reparsed
The controlled metadata inputs
Three inputs contained known synthetic EXIF camera, GPS and XMP fields. The conversion path recorded the original dimensions and metadata, generated new outputs and passed those outputs through the viewer.
Synthetic values make the privacy result repeatable without storing a person’s coordinates or device identity. Real browser behavior still depends on codec implementation, so the environment is listed beside the evidence.
What the before-and-after parse showed
The ordinary camera, GPS and XMP blocks were not copied. Output dimensions matched the requested result and the pixel encoding was new. This supports a precise statement about AnalyzeImage’s browser conversion path.
It does not support the claim that every converter removes metadata. Native applications can deliberately copy metadata, and server tools may preserve profiles or provenance according to different settings.
Color, orientation and transparency need explicit checks
Canvas decoding normally presents an orientation-corrected visible image, but output behavior should still be visually checked. Color profiles may not survive the canvas path, and JPEG needs a matte where the source is transparent.
Use a color-managed application for critical work. For privacy delivery, compare the output’s appearance and run the metadata viewer before sharing.
Conversion is not complete anonymization
A sign, face, map, document or location visible in the pixels remains. Image dimensions, file size and a newly generated timestamp may also reveal workflow context even when source EXIF is gone.
Use conversion as one technical layer. Redact visible content, select a neutral filename and consider the recipient and platform.
When a metadata cleaner is the better tool
Choose cleaning when you want the same JPEG, PNG or WebP payload and only need supported privacy fields removed. Choose conversion for new dimensions, format or compression.
Keep both outputs separate from the master and label them clearly. The right choice follows the user’s goal, not an assumption that one operation is always safer.
Repeat the test with a transparent audit trail
Use synthetic values such as ExampleCam, a clearly non-personal coordinate and a test copyright name. Record the input metadata table and hash, then convert at unchanged dimensions before trying any resize.
Save the returned MIME, detected format, output hash, dimensions and metadata table. A second run with resizing isolates whether any difference belongs to format conversion or the size change. This method makes the conclusion reproducible without exposing a real person’s data.
Keep the fixture and result manifest together so future browser releases can be compared against the same inputs.
- Synthetic metadata only
- One variable per test branch
- Input and output hashes
- Verified MIME and signature
- Second metadata parse
- Document the browser version
What this does not prove
A metadata result describes the fields and structures that the supported parser could read. It does not, by itself, prove authenticity, intent, authorship or the truth of the visible scene.