Detect EXIF, GPS, XMP, IPTC, C2PA and common AI-related metadata, then rebuild a fresh image. Processing happens in memory and no copy is kept.
The image is decoded and rebuilt as a new file instead of copying the original metadata blocks.
Camera model, timestamps, exposure information and common EXIF data are not copied into the rebuilt image.
Location-related metadata is not preserved when the image is decoded and re-encoded.
Common XMP, IPTC and Photoshop metadata blocks are not copied into the rebuilt image.
The analyzer checks for recognizable AI-related metadata strings before and after processing.
C2PA and JUMBF-related markers are checked when recognizable signatures are present.
Your image is processed entirely in memory. Nothing is written to the server and nothing is kept after the response.
What FastRemove looks for inside an image file, and what it deliberately leaves alone.
Generative image tools often write their own name, the model version and sometimes the prompt into the file itself. Those strings sit in EXIF or XMP blocks, and systems such as Midjourney, DALL·E, Stable Diffusion, Firefly and Gemini all leave recognisable traces. FastRemove reports every tag it finds in your upload and carries none of them into the result, because the new image is encoded from decoded pixels instead of being copied byte for byte.
A photo carries far more than pixels. EXIF holds the camera model, lens, exposure settings and capture time; GPS blocks hold coordinates; IPTC and XMP hold captions, copyright notices and editing history. Rather than deleting each field one by one, FastRemove decodes the image and encodes an entirely new file, so the metadata containers simply have nowhere to travel to. The output is then scanned again and the before/after report shows exactly what changed.
C2PA — usually presented as Content Credentials — stores a signed provenance manifest describing how an image was produced and edited. It is embedded as a JUMBF container, which makes it metadata rather than pixel data, so rebuilding the image does not preserve it. FastRemove flags a C2PA manifest when it recognises one and tells you whether any signature is still present in the output.
Metadata and watermarking are not the same thing. Invisible watermarks such as SynthID are encoded into the pixels themselves, survive re-encoding, and FastRemove neither targets nor removes them. A visible logo painted onto a picture also stays. Cleaning metadata is a privacy step: it does not make a synthetic image indistinguishable from a photograph and must not be used to pass one off as the other.
Image metadata is additional information stored inside an image file. Depending on the camera, smartphone or application that created the image, this information can include the device model, capture date, location, software and other technical properties.
An image can look completely normal while containing information that is invisible when the image is displayed. Smartphone photographs are a common example because some files can contain location information.
Some generative image systems add metadata describing the software or generation process. Other systems may use provenance technologies such as C2PA. FastRemove checks for recognizable strings associated with several systems. This is a heuristic detector, not a forensic AI classifier.
No. Metadata can be removed, modified or added by software. The presence of a particular marker can provide evidence about an image's processing history, but it is not absolute proof of its origin.
Metadata and pixel-level watermarking are different technologies. Rebuilding an image can remove metadata, but it does not remove an invisible watermark encoded into the actual pixels, and FastRemove does not attempt to.
FastRemove uses PHP GD to decode the uploaded image and encode a new image. The output is generated from the decoded pixels instead of copying the original file byte-for-byte.
No. The upload is read into memory, processed and returned in the same response. Neither the original nor the cleaned image is written to disk.
Common questions about image metadata cleaning.
No. This version uses PHP GD. JPEG, PNG and WebP processing can be handled directly by GD when the corresponding support is enabled.
The original image is decoded and a new image is encoded. The original EXIF block is therefore not copied into the new file.
GPS information stored as image metadata is not intentionally copied into the rebuilt image. The result is also scanned for recognizable GPS markers.
The image is rebuilt instead of preserving the original metadata containers. FastRemove then scans the resulting file for recognizable C2PA/JUMBF signatures.
No. If a watermark-related string exists in metadata, rebuilding removes that metadata. Invisible watermarks encoded into the pixels themselves are a different technology and FastRemove does not target them.
No. The AI detector is based on recognizable metadata markers. A marker can suggest a generator, but it is not definitive proof of the image's origin.
No. The file is processed in memory during the request and discarded when the response is sent. Nothing is written to disk.
JPEG, PNG and WebP are supported. TIFF, HEIC and other formats require additional image-processing libraries.
JPEG is re-encoded at a high quality setting, so there can be a small amount of additional compression. PNG uses lossless encoding for its pixel data.
No. FastRemove detects known metadata and textual signatures. It cannot determine with certainty whether an image was created by a human, camera, editing application or generative AI.