about
a lightweight, browser-based tool for quick image authenticity checks. no uploads, no ai, no magic — just practical forensic heuristics.
what this tool does
real pls runs a series of forensic checks on images to help identify potential signs of manipulation. these checks analyze metadata, compression artifacts, and pixel-level patterns that may indicate editing.
all processing happens entirely in your browser. your images never leave your device, and there's no server-side component involved in the analysis.
what this tool does not do
this is not a lie detector for images.
real pls cannot definitively prove whether an image is "real" or "fake." no tool can do this with certainty. the checks here provide signals that may warrant further investigation, but they are not conclusive evidence of manipulation.
- no ai/ml detection — we don't use machine learning models
- no deepfake detection — cannot identify synthetic faces
- not forensically admissible — results should not be used as legal evidence
- no absolute certainty — every check has limitations
the checks
file type detection
verifies the actual file format by reading magic bytes and comparing with the declared type and extension.
a mismatch doesn't necessarily indicate manipulation.
exif metadata analysis
extracts and analyzes exif data including camera make/model, software tags, timestamps, and gps coordinates. flags known editing software signatures.
exif can be easily modified or stripped. missing exif is common after social media sharing.
jpeg quality estimation
analyzes the jpeg quantization tables to estimate the compression quality level.
this is a rough estimate. low quality doesn't indicate manipulation.
noise consistency analysis
analyzes texture/noise patterns across the image. regions with significantly different noise characteristics may indicate compositing.
natural images often have varying noise levels. high sensitivity leads to false positives.
clone detection
uses block hashing to find repeated regions that are spatially separated.
patterns and textures cause false positives. cannot detect all types of cloning.
error level analysis (ela)
re-encodes the image as jpeg and compares error levels. areas modified more recently may show different error levels.
ela is notorious for misleading results. should never be used as evidence of manipulation.
best practices
- use multiple signals together — no single check is conclusive
- consider the image's history and context
- sophisticated edits may evade all checks
- when in doubt, consult a professional
- never make accusations based solely on these results