Grabtext — Copy Text From Any Image or Video Frame (OCR)

Copy the text out of any image. Right-click a screenshot, a scan, a photo of a receipt — and the words are on your clipboard a second later, selectable and editable. Chrome can't do this. Everything else that can wants you to upload the picture to a website first. Grabtext reads it on your own computer and makes no network requests at all. WHAT YOU CAN GRAB • Any image on any page — right-click → "Copy text from this image" • Any video frame — pause and right-click → "Copy text from this video frame". The command on a tutorial slide, the number in a screen recording, a subtitle. The frame is read at the video's own resolution, not off your screen. • Any part of the page — press Alt+Shift+T and drag a box. This is the one that covers everything else: charts drawn on a canvas, map labels, CSS background images, scanned PDFs, a screenshot someone pasted into a chat. WHAT COMES BACK The text lands in a small panel, and it is editable before you copy it — OCR reads 95% of a screenshot perfectly and gets one serial number wrong, and fixing that one character beats retyping the lot. • "Copy" takes it as it looks. "One line" strips the line breaks, for a key, a URL or a serial number that wrapped. • Column spacing is preserved, so a screenshotted table pastes as a table. • Dark screenshots are inverted automatically. Light text on a dark ground is what terminals, code editors and subtitles all look like, and it is the case most readers do worst on. • Small images are scaled up before reading, which is why a 200-pixel thumbnail still comes out legible. • Optional: join wrapped lines into paragraphs, and convert curly quotes to straight ones for pasting into a terminal or code editor. EIGHT LANGUAGES, ALL INCLUDED English, Spanish, French, German, Portuguese, Italian, Russian and Simplified Chinese. Every language pack ships inside the extension — switching never downloads anything, and the picker in the result panel re-reads the same image without capturing it again. PRIVACY, IN PLAIN TERMS • No network requests of any kind. No servers, no accounts, no analytics. • No host permissions. Grabtext does not ask to "read and change your data on all websites", because it does not need to. It uses activeTab: one tab, one action you started, and only for as long as that action takes. • The recogniser — Tesseract, open source — is bundled, about 25 MB of it, precisely so your screenshots never have to leave the machine. • The last 20 grabs are kept on this computer so you can copy one again. Text and hostname only: never the image, never the page address. One switch turns it off and deletes what is already there. FREE No account, no quota, no "5 free scans a day". There is no per-scan cost to pass on. KNOWN LIMITS, SO YOU DON'T HAVE TO FIND OUT • DRM-protected video (Netflix, Disney+ and similar) hands out blank frames to everything, including Chrome's own screenshot. Grabtext detects this and says so instead of returning nothing. • Handwriting is not supported. Tesseract reads printed text. • Chrome does not allow extensions to run on chrome:// pages or the Web Store, so nothing can be grabbed there. • The first grab after starting your browser takes a second or two longer while the recogniser loads. After that it stays warm.