Upload the crop, keep the whole picture

With the part they never had — and the people here are honest about how far that proof goes. Seven separate conversations crop a small piece off an image before uploading it and keep the uncropped file; if the picture is stolen, the thief’s copy is missing the edge only you can produce, which is the same logic galleries use when they hide part of a painting behind the frame, one account says. Then the limits. Two conversations say the tip is ironic or impractical — a thief can put up a less-cropped version if they got yours from somewhere else, or crop yours further so the comparison fails; one exchange, none against one, and one caution say a watermark beats a crop because it travels with the image; one caution says the method only proves anything if the cropped-out area is distinctive; one says thieves upscale to fake an uncropped version; and two say that sharing your proof image to win the argument hands the thief a fuller copy — the process is circular. Two cautions settle where the accounts land: the crop is enough in most ordinary disputes, and it is not security. One account takes it to the absurd — upload a single pixel — to show the logic has a bottom. Two conversations say the simpler proof is often the date: your upload predates theirs — with one caution that this fails if you lose the account or the thief gets your upload removed. One account says the practical remedy against a repost is a takedown request to the platform rather than a lawyer, which in the US means a copyright notice under the takedown law; how to file one, and what to do when the thief is abroad or anonymous, are gaps the accounts named — and two accounts say those cases are common and usually unactionable. So: crop and upload, keep the original, note the date, and read the next question for the proof that actually holds.

The file the internet never sees. Three separate conversations say the strongest evidence of authorship is the original source: the camera’s raw data, the layered project file with every stroke on its own layer, the intermediate process shots — the things a thief who only has your finished export cannot produce; two conversations say a layered file is definitive, two say the raw file beats the crop, and one says the same of saving drafts for written work. One exchange, two accounts to two, argues whether keeping raw files is necessary or overkill for a hobbyist — two cautions say the storage is real, one says phone raw files add little — and the accounts’ decider is whether you would ever need to prove it. The corrections tidy the technical record: raw is not an acronym; a raw file is unprocessed sensor data, not a big picture — one exchange, one account to two, and two corrections say it cannot be viewed without software that converts it; and it is lossless sensor data rather than merely uncompressed. Two conversations say serious photographers shoot raw anyway because it keeps the shadow detail a compressed file throws away, and one exchange, one account to two, has one saying you are not a photographer without it and the two that a phone photographer is a photographer — this page takes no side. Single accounts add the low-tech versions: photograph the physical work from a second angle and keep the physical original; keep drafts and work-in-progress for essays; one puts image files under a version-control system so every change is dated privately; and film photographers once notched the camera’s film gate so every negative from that camera carried its mark — one account calls that the ancestor of the crop.

The people here split on it, and their reasons are worth more than a rule. Two separate conversations say a watermark — visible or invisible — beats a crop because it survives a thief’s further cropping; one exchange, two accounts to two, says a unique watermark such as your handle is useful against the answer that a professional removes any watermark and that there are tools for exactly that; another, one to two, says a watermark deters casual theft against it degrading the work and putting buyers off. Single accounts on the mechanics: a transparent logo is weak because the plain high-resolution logo can be lifted and reused; a faint translucent mark across the whole image is harder to remove without wrecking the picture; never upload the high-resolution version at all, because a small file is worth less to a thief; writers hide markers in the text itself — an acrostic in the dedication, a deliberately odd sentence — that survive copying; one caution says an invisible metadata watermark is not foolproof either. One correction from the US side: a watermark is not the same as a formal copyright notice in law, though US law does restrict stripping the notice out. Two cautions from the aesthetic side: heavy marks make work look amateur and stop people sharing it, which for some accounts is the point of sharing it. The accounts’ decider, where they give one, is what you are protecting against — the casual reposter, who a watermark deters, or the determined one, who it does not — and how much of the picture you are willing to spoil to make the point.

No to the second — it is a myth — and to the first, only if you might one day sue, the people here and the public guidance agree. One correction in these conversations says copyright applies to original work automatically, with no application, and registration is an optional extra step; the US Copyright Office says exactly that, and adds that registering is what lets you sue for infringement and claim statutory damages there — in the UK there is no register at all and the right is simply yours from creation. One correction ends an old myth: mailing a sealed copy to yourself provides no legal protection whatsoever, and the US Copyright Office says the same. Two conversations say formal registration is the reliable route for writers dealing with publishers, and one caution says publishers rarely steal but individuals submit stolen work to them. The rest of the law strand is the accounts’ realism: one exchange, three accounts to one, says these proofs work in disputes against one who says they have only ever settled internet arguments; one account says proving ownership in court is prohibitively expensive and the practical tool is the takedown notice; one says chasing an anonymous or foreign thief is impractical; one says a company stealing a design for profit is a different and more actionable case; and one exchange, one each way, asks whether the whole overhead is worth it for anyone but a professional with money at stake. One caution rated high says all of this is jurisdiction-specific and none of it is legal advice. On the newer idea: three exchanges — none against one, none against three, none against one — have no account in these conversations defending tokens on a blockchain as protection; the corrections say a token does not include copyright and does not stop anyone using the image. What to do step by step after you have proved it, and how it works across platforms or for video, are gaps the accounts named.

Search the picture, then ask for one they cannot have prepared. Five separate conversations run a profile picture or a sent photo through a reverse image search — several free ones exist — and say it catches stolen photos and fake profiles in seconds; one caution says contest organisers who fail to do this get fooled by stolen entries. Three conversations, when the search is clean, ask the person for a new photo doing something specific — an unusual pose, holding a particular object — which a catfish working from a stolen set cannot supply. Three conversations turn the tool round for your own privacy: do not use the same picture on a dating app as on your other accounts, or the search leads a stranger straight to your name, your workplace and your friends — take new photos for dating. Four conversations use a generated face — a realistic person who does not exist — as the picture on an anonymous account, so that nothing links back; and the exchange about them, three accounts to four in one conversation, is whether the faces look obviously wrong or entirely real until you know — four accounts in the prevalence notes say they feel uncanny, three say the backgrounds hide half-formed second faces, and single accounts list the tells: teeth, ears, mismatched eyes, glasses that dissolve. One account says people who cannot recognise faces do not get the creeping feeling at all. The contraindication is in the safety line: not for marketing or anything commercial — two cautions rated high say the copyright likely belongs to whoever trained the model and the law differs by country — and two exchanges, one each way and none against two, have the caution side unanswered. One last privacy line from two accounts, for the person sending a photo of a document: blurring is not enough, because software can reconstruct what is blurred — crop or black out instead, and the exchange on that, none against two, agrees.