Image Metadata Remover — Strip EXIF, GPS & Serials, Lossless

Remove the metadata that gives a photo away — GPS location, camera and phone serial numbers, names and contact details, captions and editing history — from JPEG, PNG, WebP, HEIC, AVIF and TIFF images, and embedded thumbnails from JPEG, PNG and WebP, without re-encoding them: the pixels stay identical, and every file is checked before it is returned. It is for anyone publishing, sending or selling photos, and for AI agents, which can call and pay for it through Apify's MCP server and agentic payments; Content Credentials (C2PA) and AI-generation labels are always kept, never stripped. It costs $0.001 per image, and an image that fails is never charged.

Example result

Example record for the minimal input under How to use it, as this Actor's code returned it in a local run on 2026-10-27 (the cleaned 835-byte JPEG is in the run's key-value store; outputUrl is shown in its platform form, the store id written as STORE_ID). The record names the fields it found — never their values:

{
  "url": "https://raw.githubusercontent.com/python-pillow/Pillow/262e8e48229197ef08792de41a6bdd94e42219a0/Tests/images/exif_gps.jpg",
  "ok": true,
  "format": "jpeg",
  "width": 8,
  "height": 8,
  "reportOnly": false,
  "removedGroups": [
    "location",
    "device",
    "thumbnails",
    "otherMetadata"
  ],
  "keptGroups": [
    "timestamps",
    "structure"
  ],
  "fieldsRemoved": 17,
  "fieldsKept": 1,
  "found": {
    "location": [
      "GPS:GPSAltitudeRef",
      "GPS:GPSDateStamp",
      "GPS:GPSDifferential",
      "GPS:GPSLatitude",
      "GPS:GPSVersionID"
    ],
    "device": [
      "ExifIFD:LensInfo",
      "ExifIFD:LensMake",
      "IFD0:Make",
      "IFD0:Model",
      "IFD0:Software"
    ],
    "timestamps": [
      "ExifIFD:DateTimeOriginal"
    ],
    "thumbnails": [
      "IFD0:OtherImageLength"
    ],
    "otherMetadata": [
      "ExifIFD:ExposureCompensation",
      "ExifIFD:ExposureTime",
      "ExifIFD:ISOSpeed",
      "ExifIFD:Opto-ElectricConvFactor",
      "ExifIFD:Sharpness",
      "IFD0:BlackLevelDeltaH"
    ]
  },
  "aiLabel": false,
  "digitalSourceTypes": [],
  "contentCredentials": null,
  "bytesBefore": 1097,
  "bytesAfter": 835,
  "pixelsIdentical": true,
  "notes": [],
  "key": "cleaned-000-81d9594089.jpg",
  "outputUrl": "https://api.apify.com/v2/key-value-stores/STORE_ID/records/cleaned-000-81d9594089.jpg",
  "sha256": "1c0ad852a7d2944f3decdbfe8b9bad08f9c6263296926740c98ce42bdb1bb90b"
}

How to use it

Start a run in Apify Console, by API or from an agent with an input like this one — it is the input Apify runs every day to test this Actor:

{"urls": ["https://raw.githubusercontent.com/python-pillow/Pillow/262e8e48229197ef08792de41a6bdd94e42219a0/Tests/images/exif_gps.jpg"]}

Field	Type	Required	Default	Meaning
urls	array of strings	yes	—	Image URLs (JPEG, PNG, WebP, HEIC/HEIF, AVIF, TIFF), http or https, 1–100 per run, up to 100 MB each. Each URL becomes one record and, unless reportOnly is set, one cleaned file.
removeLocation	boolean	no	true	GPS position, altitude and direction, and place names (city, state, country, sub-location) in EXIF, XMP and IPTC.
removeDevice	boolean	no	true	Camera and phone make, model and serial numbers, lens details and serial, owner name, unique image ID, the maker's private notes (MakerNotes), and the software and computer that wrote the file.
removePeople	boolean	no	true	Artist, creator, author and by-line names, the creator's address, phone, email and website, people shown, and the AI prompt writer's name.
removeTimestamps	boolean	no	false	When the photo was taken, digitized and modified, with time-zone offsets and sub-seconds, in EXIF, XMP, IPTC and PNG. Kept unless set.
removeThumbnails	boolean	no	true	Embedded thumbnails and preview images in JPEG, PNG and WebP (EXIF, Photoshop, XMP, MPF previews) and anything appended after a JPEG's end (a motion-photo video, an appended image). A thumbnail can show the photo before it was cropped or edited. A HEIF or AVIF thumbnail item and a TIFF's reduced-resolution image are kept.
removeDescriptions	boolean	no	true	Titles, captions, descriptions, keywords, comments (EXIF, XMP, IPTC, JPEG and PNG comments) and ratings.
removeEditingHistory	boolean	no	true	XMP document and instance IDs, the editing history (which software did what, sometimes with file paths), Lightroom and Camera Raw settings, and Photoshop layers.
removeRights	boolean	no	false	Copyright notices, rights, usage terms, licence and credit lines. Kept unless set: they are rights-management information.
removeColorProfile	boolean	no	false	The embedded ICC colour profile. Kept unless set: removing it can change how colours look.
removeOtherMetadata	boolean	no	true	Everything else not listed: exposure, flash, focus and scene settings, lens corrections, unknown tags, Photoshop print settings.
reportOnly	boolean	no	false	List what each image carries, group by group, without changing or storing it. Charged like a cleaned image.

Every limit above is enforced by the input schema and again in code. The groups, in more detail:

• Always kept, with no switch: provenance — Content Credentials (the C2PA manifest store: JPEG APP11, PNG caBX, WebP C2PA, the HEIF/AVIF box), the IPTC digital source type and AI-system fields in XMP, the remote-manifest link, and the text an image generator writes into a PNG (AUTOMATIC1111 and Forge parameters, ComfyUI prompt and workflow, InvokeAI, Fooocus, NovelAI); and structure — orientation, resolution, colour space, the coded image's own parameters, an MPF stereo pair and an HDR gain map.
• A field ExifTool reads but that no group above names (exposure, flash, scene, unknown tags, a Print Image Matching block) is otherMetadata, and so is an APPn segment of a JPEG outside the standard set (JFIF, EXIF, XMP, ICC, MPF, Content Credentials, Photoshop/IPTC, Adobe): a maker's private block that the Actor cuts out itself. Face and person regions (Lightroom, digiKam, Picasa, Microsoft Photo, IPTC image regions) are people; drone flight altitudes and attitude are location; ImageMagick's exif: PNG keywords follow the EXIF field they copy. Every field lands in exactly one group.

Events and prices

Pay per event; platform usage is included in the price, never passed through.

Event	Title	Charged when	FREE / BRONZE	SILVER	GOLD / PLATINUM / DIAMOND
image-processed	Image processed	One image cleaned and stored, or with reportOnly read and reported.	$0.001	$0.0009	$0.0008

One event per image, charged only after the cleaned file is stored (or the report written) and its record written; whatever the format, size and number of fields removed. Apify also charges its standard apify-actor-start event: $0.00005 per GB of run memory (1 event, $0.00005, at the default 1,024 MB). The minimum maximum-charge a caller can set is $0.01.

Worked example on the FREE tier: the photo above costs $0.001 plus $0.00005 for the start, $0.00105 in all; a run of 100 photos costs 100 × $0.001 = $0.10 plus the start. Nothing is charged for an image that fails.

Call it from an agent

Through Apify's MCP server (https://mcp.apify.com), tool call-actor:

{"actor": "hermeskd/image-metadata-remover", "input": {"urls": ["https://raw.githubusercontent.com/python-pillow/Pillow/262e8e48229197ef08792de41a6bdd94e42219a0/Tests/images/exif_gps.jpg"]}, "callOptions": {"maxTotalChargeUsd": 0.01, "timeout": 1800}, "waitSecs": 30}

The call returns the run's status and its defaultDatasetId. Read the records with get-dataset-items:

{"datasetId": "<defaultDatasetId from the call>", "limit": 100}

Each record's outputUrl is the cleaned image itself (also readable with get-key-value-store-record and the record's key). As a dedicated tool, add https://mcp.apify.com?tools=hermeskd/image-metadata-remover to your MCP client. A photo finishes in about a second of work, so one call returns the result inside the 30-second wait; maxTotalChargeUsd caps what the run can charge, and when it is reached the remaining URLs come back as charge_limit_reached records and are not charged. To see what a photo reveals without changing it, add "reportOnly": true.

Output fields

One record per input URL, and one cleaned image per successful record in the run's default key-value store (none with reportOnly).

Field	Meaning
url	The input URL; null only on the one invalid_input record of a run given no input.
ok	true when the image was cleaned and stored (or, with reportOnly, read); false with error.
format	The image format read from the file's bytes; the cleaned file keeps it.
width	Width in pixels as stored (before EXIF orientation is applied).
height	Height in pixels as stored.
reportOnly	true when the run only listed the metadata; nothing was changed or stored.
removedGroups	The field groups found in the image and removed from it.
keptGroups	The field groups found in the image and kept: the ones switched off, plus provenance (Content Credentials and AI labels, always kept) and structure (orientation, resolution, colour space).
fieldsRemoved	How many metadata fields were removed.
fieldsKept	How many metadata fields were kept, structure fields not counted.
found	Every metadata field found, by group: field names as directory:name (GPS:GPSLatitude, IFD0:Make, XMP-dc:Creator). Names only — values are never returned. structure is not listed.
aiLabel	true when the image carries a label saying it was generated by AI (an IPTC digital source type such as trainedAlgorithmicMedia, or an image generator's PNG text). Always kept.
digitalSourceTypes	IPTC digital source type codes found in the Content Credentials or XMP, such as trainedAlgorithmicMedia or digitalCapture.
contentCredentials	null when the image has none; otherwise {present, kept (always true), stateBefore and stateAfter (Trusted, Valid or Invalid), verifiesAfterEdit, generator, digitalSourceTypes, carriesPersonalData, personalAssertions}. Removing any metadata makes a credential fail verification: it stays in the file, showing the file was edited after signing. carriesPersonalData is true when the kept credential holds an assertion that can name a person or place (an author or creator, an identity, embedded EXIF or IPTC); personalAssertions lists those assertions' labels, never their values. The Actor never removes them.
bytesBefore	Size of the downloaded image.
bytesAfter	Size of the cleaned image.
pixelsIdentical	true: the coded image data (JPEG, PNG, WebP) or the decoded pixels (TIFF, HEIF, AVIF) were checked equal after the rewrite. A file that fails the check is never returned.
notes	Anything kept that a caller may not expect (bytes after the end of a JPEG, a HEIF thumbnail image).
key	Key of the cleaned image in the run's default key-value store.
outputUrl	The cleaned image's key-value store record URL (API).
sha256	SHA-256 of the cleaned file.
error	{code, message} when ok is false.

Limits and run time

• 1–100 URLs per run; files up to 100 MB each. The format is read from the file's bytes, never from its name or the server's content type; GIF, BMP, SVG, raw camera files and PDF are refused as not_image.
• Nothing is re-encoded: ExifTool rewrites only the metadata. Before a file is returned, the Actor checks that every field of every removed group is gone, that every kept field is still there, that the coded image (JPEG, PNG, WebP) is byte-for-byte unchanged or the decoded pixels of every frame (TIFF, HEIF, AVIF) are equal, and that Content Credentials are carried byte-for-byte. A file that fails any check is never returned. TIFF, HEIF and AVIF images over 100 megapixels across their frames are refused as too_large, because their check decodes them. A TIFF, HEIF or AVIF image the decoder cannot read is not_image, found before anything is written. Every image build of this Actor runs the same checks on planted images of all six formats and on Content Credentials fixtures where its ExifTool is installed, and a build where any check fails is never deployed; each run logs the ExifTool version.
• Content Credentials: a credential signs every byte of the file outside itself, so removing any metadata from a signed image makes the credential fail verification. It stays in the file, showing the file was edited after signing, and the record says so (contentCredentials.verifiesAfterEdit: false). With every switch off nothing is removed, and the file comes back byte for byte with its credential still verifying. Where a rewrite would drop the credential (a TIFF that carries one, with the ExifTool release this Actor was tested on), the image is refused as metadata_not_removable rather than returned without it. A kept credential can name its author or carry other personal assertions (a CreativeWork author, an identity, embedded EXIF or IPTC); the Actor does not remove it, and the record says so in contentCredentials.carriesPersonalData, naming the assertions by label. A damaged credential that reads differently once the file around it changes is refused as metadata_not_removable, so it is never altered.
• A metadata block the Actor finds in the file but ExifTool cannot read (a PNG whose EXIF chunk follows a Content Credentials chunk) is refused as metadata_not_removable, never passed over as empty.
• Thumbnails and previews are removed from JPEG, PNG and WebP. A HEIF or AVIF thumbnail is its own image item, not metadata, and a TIFF's second and later images are its pages or reduced-resolution copies: both are kept, and the record's notes say so. Bytes after the end of a JPEG (a motion-photo video, an appended image or preview) go with removeThumbnails; kept, they are noted. An MPF stereo pair and an HDR gain map are part of the picture and always kept. A Print Image Matching block inside a camera maker's notes is kept with them when removeDevice is off.
• Removing the colour profile can change how colours look; the copyright notice, the licence and the dates are kept unless you switch them off.
• Downloads: files download at 2 MB/s or faster — each download has a budget of 30 s plus 1 s per 2 MB of its declared size (or of the 100 MB cap when the server declares none), and stops when no byte arrives for 30 s; at most 2 requests a second go to one host. GitHub release links are refused (github.com's robots.txt disallows them); give a raw.githubusercontent.com address or your own storage.
• Typical run time, measured on 2026-10-27 on one CPU core: about 0.6 to 1 second of work per photo; the 15 benchmark images, from 1 KB to a 51-megapixel HEIC, took 12.9 seconds in all. The default 1,024 MB is enough for a 51-megapixel HEIC (569 MB at peak); the default run timeout is 1,800 seconds.

Errors

Errors are records with ok: false and error.code; none is charged and no file is stored.

error.code	Meaning
download_failed	The URL did not return the file (status, network error, a stall of 30 s, or slower than its download budget).
not_image	Not a JPEG, PNG, WebP, HEIC/HEIF, AVIF or TIFF image (read from its bytes), an empty body, or a damaged file the container walk, ExifTool or the decoder cannot read.
too_large	Over 100 MB, or a TIFF, HEIF or AVIF image over 100 megapixels across its frames.
metadata_not_removable	A field could not be removed without rewriting the image, a metadata block could not be read, the metadata is too damaged to rewrite, removing it would also remove the Content Credentials or AI label, or a damaged credential reads differently after the edit; the message names the reason.
robots_disallowed	robots.txt forbids fetching the URL.
blocked_host	The host is on the deny-list, or resolves to a private or local address.
timeout	The image did not finish within the run's time budget.
internal_error	Unexpected failure, including a rewrite that failed the Actor's own checks; message carries the class.
charge_limit_reached	The run's maximum total charge cannot pay for this image; it was not processed.
invalid_input	The run input passed the input schema but cannot be used (for example a switch that is not true or false, or no URL at all); the message names the field. One record per input given, or one with url: null when none was given; nothing is processed.

The run fails only when every input failed with an Actor-side code (internal_error, timeout), so a broken Actor is visible. Every other run succeeds — including one where nothing was cleaned because of the caller's own inputs, or where those and Actor-side failures are mixed.

Sources and terms

The source is the caller's image. Process only images you have the right to process. The Actor removes metadata; it never removes Content Credentials, AI-generation labels, watermarks or signatures, and it is not a way to pass an AI image off as a photograph. Its records never contain a metadata value — not a position, serial number, name, address or caption, and not the author or signer named inside a credential — only field names, counts, the credential's verification state, its generator software, its IPTC digital source type and the labels of its personal assertions. A kept Content Credential can name its author, and the Actor does not remove it. The cleaned image is returned to the caller in the run's own key-value store. Fetching follows the Hermes fetch policy: robots.txt is obeyed for the user-agent HermesBot (and *), and a disallowed URL is refused without a request; platforms whose terms forbid automated access, and people-search sites, are on a deny-list and refused without a request; private and local addresses are refused, after DNS and on every redirect; the user-agent is HermesBot/1.0 (+https://apify.com/hermeskd); no login, cookies, credentials, CAPTCHA solving, evasion or proxies are used.

The tools are ExifTool by Phil Harvey (the Debian package, Perl Artistic Licence or GPL), c2pa-python from the Content Authenticity Initiative's C2PA project (Apache-2.0 or MIT) to read Content Credentials, and Pillow with pillow-heif (HPND; BSD-3-Clause, with libheif under LGPL) to decode images for the pixel check.

Not affiliated

This Actor is not affiliated with ExifTool or its author, with the C2PA, the Content Authenticity Initiative or Adobe, with the Pillow project, whose public test images the examples use, with the pillow-heif project, or with IPTC. ExifTool, c2pa-python, Pillow and pillow-heif are used under their open-source licences.

Related Hermes Actors

• Article Extractor — Clean Text, Title & Date, $0.001 an Article — an article's clean text, title and publish date, without the rest of the page.
• Career Site Job Search — Verified Greenhouse & Lever Boards — search open jobs in the Hermes directory of Greenhouse and Lever boards, each confirmed live.
• Speech to Text — Transcribe & Translate to English, SRT/VTT — speech to text with SRT and VTT subtitles.
• Video Converter & MP4 to MP3 — Trim, GIF; Audio $0.001 a Minute — audio and video conversion: MP4 to MP3, trim, GIF.
• Broken Link Checker That Never Charges for the Broken Links — broken links, 404s and redirects on a site.
• Company ATS Finder — Which Hiring System & Job Board It Uses — which hiring system and job board a company uses.
• DNS Lookup & SPF, DKIM, DMARC Check — Findings by Severity — DNS records and SPF, DKIM and DMARC checks.
• Word, Excel & PowerPoint to Markdown — $0.002 per Document — Word, Excel, PowerPoint and other office files to Markdown.
• Domain Availability Checker — Registry RDAP, No Contact Data — whether domain names are available, from registry RDAP.
• File Downloader — Save Files & Images from URLs, $0.001 a File — files and images from URLs, saved with a public link each.
• Greenhouse Job Board Scraper — Open Jobs, Live, Any Company — every open job on a company's Greenhouse board.
• Image to Text OCR — Lines & Words with Boxes, $0.002 an Image — OCR: the text in images.
• Image Upscaler & Photo Restorer — AI 2x/4x, Agent-Payable — image upscaling 2x or 4x, and photo restoration.
• Lever Jobs Scraper — Postings from Any Lever Site, US & EU — every published job on a company's Lever site, global or EU.
• Lighthouse Audit, No API Key — Lab Core Web Vitals & SEO Scores — Lighthouse scores and Core Web Vitals.
• Page Change Monitor — Detect Website Changes, $0.001 a Check — what changed on a page since the last check.
• PDF to Markdown & PDF to Text — PDF Text Extractor with OCR — PDF to Markdown and text, with OCR for scanned pages.
• Remote Jobs Scraper — Remote OK Listings, Salary Filter, Links — current remote jobs from Remote OK, with salary and listing links.
• RSS Feed Reader — RSS, Atom & JSON Feeds to JSON, $0.001 a Feed — every item of an RSS, Atom or JSON feed.
• SEO Audit — On-Page Checks, Schema & Hreflang, $0.004 a Page — an on-page SEO audit: meta tags, schema, headings and links.
• Sitemap URL Extractor — Indexes, Gzip, RSS; $0.0003 per URL — the URLs in a site's sitemaps.
• URL to Markdown for LLMs — From $0.001 a Page, Agent-Payable — a public job page as clean Markdown where the site allows it (deny-listed job sites are refused).
• Video Frame Extractor — by Scene, Interval or Timestamp — frames, thumbnails and scenes from a video.
• Website Screenshot API — Elements, HTML, Dark Mode & PDF Print — screenshots and PDF prints of web pages.
• Website Tech Stack Lookup — Live Detection, $0.001 a Site — the technologies a website runs, detected live.
• Crossref DOI Metadata — Titles, Dates & Citation Counts — crossref doi metadata.
• JSON Dataset Cleaner — Merge, Dedup & Export Rows — json dataset cleaner.
• Public HTTP GET — Response Body, Status & Hash — public http get.
• Image Format Converter — Resize & Save WebP, PNG or JPEG — image format converter.
• Website Icon Finder — Favicons, Touch Icons & Theme Colors — declared website icons, theme colors and manifest links.

Changelog

• 1.0 (2026-10-27) — before release, from the QA sweep: camera photos with a Print Image Matching block cleaned; an unreadable HEIF, AVIF or TIFF metadata block refused, never passed as empty; damaged or oversize TIFF, HEIF and AVIF files reported as not_image and too_large; a TIFF's pages kept; face regions, drone positions and ImageMagick's EXIF keywords in their groups; appended images and unknown APPn segments removed; carriesPersonalData on kept credentials; the image checks itself at build.
• 1.0 (2026-10-27) — first release: lossless removal of location, device, people, timestamp, thumbnail, description, editing-history, rights, colour-profile and other metadata from JPEG, PNG, WebP, HEIC/HEIF, AVIF and TIFF, one switch per group; Content Credentials and AI labels always kept; a report-only mode; every file checked before it is returned.