FFT Spectrum Analyzer
Transform images into the frequency domain using a 2D Fast Fourier Transform. AI-generated images, GAN artifacts, and compositing patterns reveal themselves as anomalies invisible in the pixel domain.
1,200+ spectrums analyzed
Drag & drop a file here, Ctrl+V to paste, oror
AI & GAN Detection
GAN-generated images often produce characteristic periodic patterns in the frequency domain — grid-like artifacts from the generator's upsampling layers. These are invisible in the pixel domain but clearly visible in the FFT spectrum.
Spectral Analysis
The tool computes a high-frequency ratio (energy in outer vs inner spectrum) and counts spectral peaks. High HF ratio or multiple peaks suggest artificial processing. Natural photos have smooth, monotonically decreasing radial profiles.
Web Worker Performance
The 2D FFT runs in a dedicated Web Worker thread using the Cooley-Tukey radix-2 algorithm. Images are padded to the next power of 2 and windowed (Hanning) to reduce spectral leakage. Processing time depends on image size.
What the FFT spectrum reveals
The Fast Fourier Transform converts an image from the spatial domain (pixels) into the frequency domain, revealing periodic structures that are invisible to the naked eye. Natural photographs have a characteristic frequency profile — energy concentrated at low frequencies with a smooth falloff toward high frequencies. AI-generated images, screen captures, and composites often introduce periodic artifacts that appear as bright spots or lines in the spectrum. For pixel-level analysis in the spatial domain, combine with the ELA Scanner and Noise Analysis.
How GANs leave frequency fingerprints
Generative Adversarial Networks (GANs) use transposed convolution (upsampling) layers that introduce checkerboard-like artifacts at specific frequencies. These show up as periodic bright dots in the FFT spectrum, even when the generated image looks photorealistic to the human eye. Newer diffusion models produce subtler artifacts, but frequency analysis remains one of the few methods that can detect them without a trained classifier. For AI detection using other methods, try the AI Detector or the Authenticity Checker.
JPEG compression in the frequency domain
JPEG compression works by dividing the image into 8×8 pixel blocks and applying a Discrete Cosine Transform (related to FFT) to each block. This creates characteristic cross-shaped artifacts in the spectrum at multiples of the block frequency. Double-compressed JPEGs (saved twice at different quality levels) show additional spectral features that indicate editing. The JPEG Ghost Scanner specifically targets these compression inconsistencies. For checking compression artifacts region by region, see the Clone Detection tool.
Reading the spectral metrics
The high-frequency ratio measures the proportion of spectral energy in the outer ring versus the inner region. Natural photos typically have a low HF ratio (most energy at low frequencies). A high ratio suggests artificial sharpening, AI upscaling, or noise injection. Spectral peaks count isolated bright spots in the radial profile, which indicate periodic patterns. Zero peaks is normal for natural photos. For checking lens-related artifacts that also leave frequency signatures, use the Chromatic Aberration Analyzer. Explore all forensic tools.
Frequency domain analysis is one of the most powerful techniques in image forensics because manipulations that are invisible in the spatial domain often leave clear signatures in the spectrum. When an image region is copied and pasted from a different source, the spliced area frequently has a different noise profile and compression history, creating detectable frequency discontinuities. Screen captures of AI-generated images carry the display's pixel grid frequency. Rescaled images show interpolation artifacts as spectral peaks at the resampling frequency. These patterns are difficult to remove without introducing new artifacts, making FFT analysis particularly resistant to anti-forensic techniques.
For a comprehensive image verification workflow, start with the FFT spectrum to look for obvious periodic anomalies, then move to Bit Plane Analysis to examine individual bit layers for hidden data or editing traces. The Luminance Gradient scanner reveals lighting inconsistencies that suggest compositing, while the DCT Viewer shows block-level compression quality variations within the image.
Professional forensic analysts combine multiple techniques because no single method is definitive. A photo can pass ELA analysis but fail FFT inspection, or vice versa. The PCA Scanner decomposes color channels to reveal hidden patterns, and the Histogram Analyzer shows statistical anomalies in the tonal distribution. Together, these tools provide the multi-signal approach that professional image verification demands. Learn more in our articles on spotting deepfake photos and image forensics guide.