What changed in the file?
SHA-256 fingerprints the complete bytes. Metadata reveals declared device tags, dimensions and GPS-field presence. Edited tags describe an export; they do not authenticate the capturing device.
SPOOFZY SIGNAL LABThe secret of
The secret of
file forensics.
Discover what’s really in your content.
Go beneath the image. Compare real results from Premium Spoofer and AI Retake through frequency, geometry and the details hidden inside every file.
SPATIAL → FREQUENCYILLUSTRATION
F(u,v)
2D SIGNAL ANALYSIS
Σ f(x,y)e⁻ⁱ²π(ux+vy) File identity & metadata Fourier & wavelet analysis Geometry & visual fidelity
01 / COMPARATIVE FILE EXAMINATIONREAL OUTPUT FILES
Two tools. Different creative goals.
Open the evidence.
Inspect the pixels, read the metadata and download the results. Every plot below is calculated from the files shown.
FINDING / STRONG TRANSFORMATION
Stronger geometry. Measured change.
The existing Stronger preset combines rotation, perspective and reframing within a 5% crop ceiling. Source similarity measures the change; encoding fidelity checks the file against the intended transformed image.
…Source SSIM · all 3 Strong outputs
Signal Lab previews the Premium Spoofer upgrade. This example uses the Strong preset: one source image, one recorded seed and all three outputs. It is not a representative benchmark.
Loading the measured report…
READING THE REPORT
Precision needs context.
A useful forensic report connects every number to a question you can actually answer.
Where did the image change?
Difference maps locate luminance changes. SSIM measures local structural similarity and PSNR measures pixel error on a stated analysis grid. Pose changes and mirroring can lower both scores.
How did texture and geometry change?
The 2D Fourier transform separates spatial frequencies. Haar wavelets separate detail across four scales. Structure-tensor eigenvalues describe edge direction; the upgrade’s homography records its projective transform.
Different files. Different evidence.
A filename extension does not tell the whole story. Colour, transparency and the encoded streams change what a valid comparison needs.
JPEG
Inspect DCT compression, quantization tables, chroma subsampling, progressive coding and colour profiles.
Trial SDR input; high-quality JPEG export. A changed hash alone says nothing about image quality.
PNG
Inspect Sample depth, alpha, palette transparency, colour chunks and per-chunk CRC integrity.
Trial 8-bit SDR, with transparency preserved and lossless PNG encoding after the deliberate transform.
WebP
Inspect Lossy VP8 versus lossless VP8L, alpha, colour profiles and container chunks.
Trial Still images. Opaque files export as JPEG; transparent files export as PNG.
MP4 / MOV
Inspect The actual codec, pixel format, colour tags, nominal/average frame rates, time base and audio streams.
Trial Short 8-bit SDR clips export as H.264 MP4. Supported audio is remuxed; three frames are sampled for image analysis.
WebM
Inspect Video and audio codecs inside the container, colour signalling and stream timing.
Trial Short 8-bit SDR clips export as H.264 MP4. Incompatible audio fails clearly instead of being silently removed.
HDR / 16-bit
Inspect PQ/HLG transfer functions, colour primaries, high bit depth and HDR metadata. PNG can carry HDR signalling too.
Trial Unsupported HDR/high-depth input is rejected. Preserving its brightness and colour requires a separately qualified processing path.
Format references: W3C PNG specification · FFprobe stream inspection. Reported tags are declarations; missing tags do not establish the original colour space.
SPOOFZY / SIGNAL LABRESEARCH PREVIEW
Try one file. Inspect everything.Three variations.
Three variations.
Evidence in every pixel.
A forensic look at what changes: file hashes, metadata, frequency spectra, wavelet detail and projective geometry. Real files. Measured results. No account required.
Free trial · 3 runs per hour per network · originals removed after processing · downloads expire after 30 minutes
01 SHA-256 + metadata02 2D Fourier spectrum03 Multiscale Haar wavelets04 Homography + SSIM
Inspect the methods, sampling limits and recipe in each downloadable JSON report. These measurements describe the files; platform reach and detection outcomes are not measured.