What changed on the site, newest first. Dates are the days a change reached uvpdftool.com. The About page says how the tools are tested before a change ships.
: Read aloud, study aids, personal data kept back from the AI, and an access code for signers
- Redact PDF looks for every kind of personal data at once: e-mail addresses, phone numbers, payment cards, Aadhaar and PAN numbers, IBANs, US Social Security and Irish PPS numbers, and dates if you ask, each checked by its own rule (a card number by the Luhn check, an Aadhaar number by its check digit), so fewer numbers that only look like one are marked. Summarize PDF, Chat with PDF and Translate PDF replace the same details, and any names you add, with markers such as ⟦PHONE 1⟧ before the text leaves your device, and put the real values back into the answer on your device. It is on unless you turn it off.
- OCR PDF reads 25 languages, sixteen more than before: Bengali, Marathi, Gujarati, Punjabi, Odia, Assamese, Urdu, Arabic, French, German, Italian, Russian, Chinese (simplified and traditional), Japanese and Korean, chosen from a list you can search. The text it adds can be found and copied in every script, including Indian words whose vowel sign is written before its letter, which searches missed before.
- PDF to Excel takes the answers out of filled forms, from one file or many, one row per file, as a workbook, a CSV file or JSON. Inspect PDF saves the files a PDF carries, one by one or in a ZIP. My details keeps your name, contact details, address and, if you choose, ID numbers in your browser only, and fills the empty fields of a form with them in Fill and sign PDF and in the editor; it never fills a field about someone else, such as a parent’s or an employer’s.
- The PDF reader reads a document aloud with your device’s own voices, sentence by sentence, showing each sentence on its page and turning the pages with it; voices that send the text to the browser’s maker are marked online. Compare PDF opens with the changes to amounts, percentages, dates and other numbers, each with the words around it. Summarize PDF makes study aids: up to 15 flashcards, a quiz of up to 8 questions and up to 15 key terms, each with its page, to download as text or as a file Anki imports.
- A sender can give each signer an access code to pass on apart from the email. Until it is typed, the link shows only the title and who asks, and the signed copy asks for it too; five wrong codes lock the link until the sender unlocks it, which sends the signer a new one. The certificate says that identity was checked by an access code.
- An account at an address another account used before can now buy; its checkout failed before. Compress PDF on our servers can also write the file for fast web view, so a browser shows the first page while the rest downloads. A new page, Status, shows each part of the service as it was checked every quarter of an hour, with 90 days of history.
- For developers: the Node.js library has a command line (uvpdftool compress report.pdf), and uvpdftool-mcp lets an AI assistant such as Claude run the tools on the PDFs in folders you name, and nowhere else.
: The developer documentation, two SDKs, and a dozen things from the audit list
- The developer pages are a documentation site: every endpoint on its own with its parameters, its answer and the request in cURL, Node.js, Python and PHP (the language you pick holds across pages), every tool with its options, webhooks with a signature check in three languages, every error, and a Postman collection and a richer OpenAPI description to generate clients from. A card on the home page and a footer link lead there. API jobs keep their files two hours by default now, and up to a day on request.
- Libraries for Node.js and Python (uvpdftool on npm and PyPI) do the whole flow in one call: create, upload, start, wait, download, with retries and the webhook check.
- A job no server picked up within ten minutes now ends with its task returned instead of waiting for ever. PDF to Word has a second converter: when LibreOffice cannot read a PDF, the engine writes the Word document itself, with headings, lists, tables and pictures. The servers hand large results to storage without holding them in memory.
- Stamp words typed once can be kept on your device, in Stamp PDF and the editor. The editor’s comment list filters by kind and author and downloads as a CSV file. Create PDF form lists the fields in the order readers tab through them, to move or sort by position. Edit properties can set the document’s language, which screen readers and search engines use. JPG to PDF no longer counts a picture it refused.
- Three new pages: Security (where files go, how long they are kept, how they are protected, how to report a problem), How we test (the weekly run over 2,500 real files with its latest figures, the planted-fault measurements of the checks, what runs on every change) and Electronic signatures (what a signature made here is under eIDAS, ESIGN and UETA and India’s IT Act, and what the audit trail records). The pricing page lists every currency.
: Compress keeps transparent pictures whole, and seven smaller things from a day of testing
- Compress PDF left a picture’s transparency mask in a form no reader could draw, so a logo, chart or photo with soft edges could come out wrong or not at all after compressing. Masks are now left as they are. If a file you compressed earlier shows such a picture badly, compress the original again.
- PDF to Word stopped waiting on a file that LibreOffice cannot import after four minutes instead of nine, and says the conversion took too long rather than that the file could not be converted. A picture that is not a picture, an empty file or a text file named .jpg, is marked in the list as it is added instead of failing the whole run. A file your device cannot hand over, such as a cloud placeholder not yet downloaded or a path Windows finds too long, says so and what to do.
- The check before you send says engine errors in your language, and on a phone with little memory keeps to files under 25 MB. A free visitor who reached the day’s count late and tried again after midnight in the same tab is no longer refused on yesterday’s count. The pricing page names the largest file the servers take for any server tool, 2 GB for Compress and Repair; the address /tools leads to the list of tools; the API takes “bearer” in any case, tells you how long to wait after a 429, and refuses a webhook that points back at this site.
: Ireland and the United States: twelve more portals and twenty-four starting lists
- Portal-ready uploads covers twelve more portals, each with a page of its own and the rules read from the authority’s page: the US visa (DS-160 and DS-260), Diversity Visa and passport renewal photos, USCIS online filing, the National Visa Center’s document uploads, USAJOBS, the ACT photo and the IRS Document Upload Tool; Irish Passport Online, the immigration service’s portal, MyWelfare and the citizenship form. The chooser groups the portals by country. A document item now takes a photograph of the document as well as a PDF, drawn no larger than a 300 DPI scan of an A4 page before it is brought under the limit; sizes read in KB or MB as the portal states them.
- The client workspace starts from thirty-six lists: twelve for Irish filings (Form 11, the PAYE return, CT1, VAT3 and the RTD, payroll, the CRO annual return and incorporation, capital gains, gifts and inheritances, RCT, a mortgage application, Help to Buy) and twelve for United States filings (Form 1040, Schedules C and E, Forms 1065, 1120-S, 1120 and 990, payroll, a new business, an ITIN, foreign accounts, a mortgage application), beside the twelve Indian ones, grouped by country.
- Bundle PDFs starts from a preset: a court paper book, a petition with annexures, a tender response or a loan file for India; an exhibit binder, an appellate appendix, a discovery production or an immigration petition packet for the United States; a book of appeal, a book of authorities, a tender response or a tribunal submission for Ireland. A new setting labels each document Exhibit A, B, C or Tab 1, 2, 3 in the order of the files, on its divider page, in the index and in the bookmarks, with the word before the label yours to choose.
: A client workspace for accountants: ask, receive, check, pack
- With Pro, the account page leads to Clients: a folder per client, started from a requirement list of what a filing needs (twelve starting lists for Indian filings, from the income-tax return to GST, company and EPF registrations, each with the file rules and its source; copy one and make it yours, share it with your team). Tick the documents to ask for and the client gets one email, with your address as the one to reply to, and a link where they upload each file from a phone without an account; reminders go on the schedule you set, with a fresh link each time.
- Each file that arrives is checked against the item’s rules (format by its bytes, size, pages, and that a PDF opens), shown with what it breaks, and accepted or sent back with a note the client sees. One click builds the pack on your device with Bundle PDFs: a cover, an index naming each requirement, the file that met it and the day it arrived, and page numbers. Marking the folder submitted ends the links and deletes the files 90 days later; deleting a client removes everything at once. The privacy policy has a section on it.
: Bundle PDFs: a court or tender bundle with a cover, an index and numbers
- A new tool takes many PDFs in the order you set, each with a title and, if you like, a section, and makes one document: a cover page, an index that names every section and document with its page or page range and links to it, a divider page before each section, bookmarks for every part with each document’s own bookmarks nested under it, and page numbers or Bates numbers stamped across the whole. The cover and the index are numbered i, ii, …; the numbering starts at 1 on the first page after them, so the numbers in the index hold however long the index is. It runs in your browser, in all nine languages, and writes titles in every script the site supports.
- With Pro, the settings are kept as templates: saved under a name, applied again with one click, and shared with your team like a saved workflow.
: Make your photo and signature fit an exam, job or bank portal
- Ten portals have a page of their own: UPSC, SSC, NEET (UG), JEE (Main), GATE, CUET (UG), UGC NET, IBPS, SBI and RRB. Each lists what the portal asks of a photograph, a signature, a thumb impression or a document (format, kilobytes, pixels, proportions, resolution), names the notice it was read from and the day, and links to report a change. The pages are in all nine languages of the site.
- On each page, drop your files: every picture is cropped to the proportions asked, drawn at the pixel size, encoded as JPG under the size limit, brought up to the minimum when the portal has one, checked against every rule and named after the portal and the item, in your browser. A PDF document is compressed to the limit and its pages counted. A card under each file says which rules it meets and, when software cannot help (too few pixels, a photograph the portal takes live), what to do instead.
- Every month each portal’s notice is read again and checked for the figures the page states; a notice that moved or changed is flagged for a fresh reading.
: A damaged file says so
- A PDF whose structure the reader had to skip parts of could stop a tool with “Something unexpected went wrong”. Such a file now says that it is damaged and cannot be read, in the browser tools and on the servers, and a server job on one is no longer retried. Repair PDF is the tool to try on it.
- In the editor, a click on one of the document’s own notes in the first moment after opening the file found nothing, because the notes were still being read. The click now waits for them and opens the note.
: Eight more tools in the lists
- HTML to PDF, Markdown to PDF, EPUB to PDF, PDF to SVG and the PDF/A validator were measured on the live servers and now appear in the tool lists, the site map and search. Each page shows what its sample file cost in size and time, and the PDF/A page what its sample report breaks.
- Summarize PDF, Chat with PDF and Translate PDF show their measured figures too, from runs on the live service: the credits, the time, the words, and the pages each answer cites.
: Files that would not open, unlock or merge now do
- Older bank statements and other files encrypted with 40-bit RC4 and no file identifier were refused as “a type of encryption that cannot be opened here”. They open now, with the identifier the file was made with: an empty one.
- Protect PDF encrypted a file’s pages and pictures but not its strings: the title, form field values, comments, bookmark names and links stayed readable without the password, and a reader given the password showed them garbled. They are encrypted now, as the standard requires.
- Unlock PDF and every tool that writes a decrypted file wrote binary identifiers (the ones InDesign leaves on each page) back without escaping them, which broke the written page tree; the file then failed the engine’s own read-back check. Decrypted strings are escaped again when written.
- Merge PDF refused files whose own bookmarks had a blank title or one over 500 characters, with a message meant for typed bookmarks. Blank titles are now named after their page, long ones are trimmed.
- PDF to Markdown left out the recognized text over scanned pages and gave a scanned letter as two pictures and nothing else; it reads that text now, as PDF to text and PDF to HTML do. A picture in a format the engine cannot read (an LZW-packed 1-bit logo, say) is left out of Markdown and HTML instead of refusing the whole file.
- Deskew PDF and Remove blank pages counted from “page 0 of 3” while measuring; they say the page they are looking at.
: Compress PDF makes drawings and outlined text small too
- Compress PDF now writes each page’s drawing instructions again: a page’s streams are joined into one, and the coordinates of paths and text positions are rounded to a thousandth of a point (finer under a magnifying transform), which is far below anything a screen or printer shows. A shape drawn thousands of times with a different sixth decimal each time was thousands of different byte strings; rounded, it is one, stored once. A 167 MB book whose Telugu text had been turned into outlines, which the tool used to leave at 166 MB, comes out under 31 MB in the browser, smaller than the strongest server tools made it.
- Text, names, inline pictures and the scale entries of transforms are copied byte for byte. TrueType fonts lose the outlines of the letters the document never uses and the tables a viewer never reads (the shaping tables of an Indian-script font are often most of it), with every glyph number kept where it was, so the text still reads and copies as before; fonts used by form fields and annotations are left untouched.
- While a file is compressed, the page says what is happening: reading page 12 of 37, shrinking picture 3 of 28, rewriting page 12 of 37, trimming the fonts, writing the file, and which attempt it is on when a target size was asked for. A book of a hundred megabytes takes minutes in the browser, and now says so instead of showing only a spinner.
- The server choice (Pro) kept Ghostscript’s result only when Poppler had counted the pages of both files; a file Poppler could not read had its smaller result thrown away and a copy of the original handed back. Pages are now counted by Poppler or, failing that, by our own engine, and only two known counts that differ discard a result.
: Fill in a form and change its comments inside the editor
- A PDF’s own form fields are filled in on the page in Edit PDF: text and text areas, check boxes, choices, drop-downs and lists, each in its box, with the field named beside the page and the file’s own value a click away. What you fill in is written into the fields as you save; a tick locks them, so they can no longer be changed.
- A signature field asks for your signature: drawn, typed or from a picture, it is placed in the field’s box.
- A note the file already carries is yours to change with a click: its words, its author, its place; underlines and strike-throughs can be deleted where they are. Saving takes the original out and puts yours in its place, and a note left as it was is left alone.
: Pages with form fields or comments show again in the editor
- A PDF that already carried form fields, comments or signatures opened in the editor with those pages black, the fields and notes floating on the dark: the layer that draws a file’s own annotations above your edits was painted opaque instead of clear. The page was intact underneath, and nothing in the file was lost; it is drawn as it should be now.
- While you type, the spelling of the text box is back beside the page: the right-hand panel names the word at the caret with what it might be, and offers Ignore, Add to my dictionary and the next mistake, with a way to the Spelling tab for the whole document.
: Summarize, Chat with and Translate PDF are on the home page
- The three AI tools, part of Pro, are listed with the other tools and marked “Pro · AI” so that what Pro offers is in plain sight: a summary that says which pages it comes from, answers drawn from the document and only from it, and a translation laid where the original was.
- Without Pro, the tool’s page says what it does and what it costs and offers the sign-in and the plans; nothing is uploaded first. With Pro, each run says how many credits it takes before it starts.
- Their pages carry no measured figures yet; those follow once the tools have been measured on the live service, as every server-side tool’s were.
: One page per job: 22 pages became choices inside the tools they belong to
- A tool page now offers its ways of working above the files, and the home page lists them under each tool: Compare PDF compares the words, how the pages look, or a copy against its original; PDF to JPG saves pages as JPG, PNG or TIFF or takes the pictures out; Add watermark puts on text, a picture or logo, or a rubber stamp; Split PDF cuts by ranges, every few pages, every page, at the bookmarks or under a size; Merge PDF takes files one after another or the fronts and backs of a one-sided scan; Organize PDF arranges pages, reverses them, or adds blank pages or another PDF’s pages; Remove pages takes the pages you pick or the blank ones it finds; Add page numbers adds numbers, a header and footer, or the numbers a reader shows; Clean PDF shows what the file carries, then removes it.
- Spell check lives in the editor: a Spelling tab in the side panel checks what you type as you type it and reads the whole document on request, with the same personal dictionary. Make PDF fillable is Create PDF form’s Find the fields, run as the file opens. The page that checked a signed copy against our record is part of Verify PDF signatures, on request. ODT, RTF, TXT and CSV to PDF were Word to PDF and Excel to PDF already.
- Every old address still works and leads to the tool with the choice made, in every language, so a saved link or a search result is not lost.
- The editor takes the whole window while a file is open, so a laptop shows more of the document; the right-hand panel holds only the selected item’s properties; and the toolbar’s buttons keep their width in every language.
: Every page fits the narrowest phones, in every language
- On phones 360 pixels wide, as many in India are, some pages could be pushed sideways, most in Malayalam and Tamil, whose words run long: a tool’s options and its button, the editor’s bottom row, a link to another tool, a long word in a title. Every page and every tool’s screen now fits, in all nine languages.
- Buttons whose words are longer than the room they have wrap onto a second line, and on a phone the editor’s bottom row keeps the page number, zoom and fit to width, without the arrows to move between pages, which scrolling does.
: Make PDF fillable: forms printed as PDFs become forms to fill in
- Make PDF fillable finds where a printed form wants its answers, its blanks of underscores, its lines to write on, its boxes to tick and to write in, and puts a field on each, named after its label; a label asking for a signature or a date makes a field of that kind. You check the fields on the page, change or add any, and save.
- Create PDF form has the same search, behind Find the fields.
- The terms of use now say plainly that what the tools find for you, the check before you send, the words a spell check flags, the fields found on a form and the differences between versions, helps your own review and does not promise that nothing was missed.
: Compare PDF by how the pages look
- Compare PDF compares the words or how the pages look: each pair of pages laid over each other, what only the older version has in red, what only the newer has in green, every changed area boxed, with a report of the pages that differ. It finds what words alone miss: a revised drawing, a moved logo, a changed chart, a new stamp.
- Compare PDF visually is a page of its own for it, listed with the other tools.
: Stamps in the editor
- Edit PDF places rubber stamps: APPROVED, DRAFT, CONFIDENTIAL, PAID, RECEIVED, COPY or your own words, with the date if you like, in red, blue, green or black. A stamp moves, resizes and turns like a picture, and is saved into the page as Stamp PDF draws it, its words text a search finds.
: Spell check a whole PDF, in British or American English
- Spell check PDF lists each word of a PDF that the dictionary does not know, once, with how often and on which pages it comes and what it might be, and opens the file in Edit PDF to correct them. It runs in your browser; nothing is uploaded.
- Spelling now checks British English beside American English, Spanish and Portuguese: British spelling, as in India and Ireland, unless your browser asks for American English. The editor and the new page share it.
- Words you add to your dictionary, such as names and terms of your trade, are no longer flagged, on the new page or in Edit PDF. They are kept in your browser only; the privacy and cookie policies list them, and Spell check PDF shows them and takes them out.
: Print from the reader, the document’s own page numbers, and work that survives a closed tab
- The PDF reader prints: each page is drawn at 200 dpi and sent to your browser’s print dialog, one page to a sheet, for files of up to 500 pages.
- The reader shows a document’s own page numbers, such as i and ii for a preface, in the page box and under the thumbnails, with the place in the file beside them, and goes to a page typed by its own number.
- Going to a page in the reader moves its pages only: the toolbar stays where it was.
- The editor keeps the changes you have not saved yet in your browser, and offers them back when you open the same file again, after a closed tab, a crash or a phone reclaiming memory. Nothing leaves your device; the changes go when you save, when you choose not to restore them, or after 30 days. The privacy and cookie policies list this, and the name you can give for comments.
: Comments in the editor
- Edit PDF now adds comments: a sticky note anywhere on a page, and words underlined or struck through, the line snapping to the text. They are saved as the comments Adobe Acrobat and other readers list, with your name on them if you give one, and every reader shows them where you put them.
- A Comments tab beside the pages lists the document’s own comments and yours, page by page; choosing one opens its page, and one of the document’s can be taken out, or kept again until you save.
- The check before you send now knows how many links an edit added and took out, so new links no longer ask for a review, and a link that went missing does.
: Seven new page tools, and pages of their own for six more
- Reverse PDF puts the pages in the opposite order, and Merge odd and even pages puts a document scanned one side at a time back together, each front before its back, the backs read last first when the stack was turned over.
- Insert pages adds blank pages, or another PDF’s pages, after any page, or puts them in place of a run of pages. Split PDF by bookmarks makes a file for each chapter or section, named after its bookmark, and Split PDF by size cuts a file into parts under the size an email or a form allows, weighing each part as it makes it.
- Stamp PDF puts a framed, slightly turned stamp on the pages you choose: APPROVED, DRAFT, CONFIDENTIAL, PAID, RECEIVED, COPY or your own words, in any script the site supports, with the date if you like. Edit page labels sets the page numbers a reader shows, such as i and ii for a preface and 1, 2 after it, without changing the pages.
- PDF to PNG and Extract images now have pages of their own, and so do ODT, RTF, TXT and CSV to PDF; those four join the list of tools once they are measured on the live service, as every page that uses our servers does.
- After a redaction, the check no longer asks for a review when a box reached into the top of the next line: the letters it reached were taken out with the rest under the box, as a redaction should.
: Check before you send
- Redact, Compress and Edit PDF now read their result back before you download it, with two readers, pdf.js and the site’s own engine, and compare it with the original: the text, the pages as pictures, the links, form fields and bookmarks, letters that can no longer be read, text pushed off the page, and what the file carries unseen. Each check says Passed, Needs review, Failed or Not tested, and a report with both files’ SHA-256 can be downloaded.
- A new page, Check before you send, does the same for any two files: the original and the version made from it. For a redaction it finds where the pages turned black and fails it when the words are still in the file underneath, as when a box was only painted over them, and it looks for any words you name.
- On a test set of 200 files, half of them given a fault on purpose, the checks found every fault and raised no false alarm; the page says what the set covers and what it does not.
: See and remove what a PDF hides, use the tools offline, and merges that keep more
- Inspect PDF lists what a file carries besides its pages: its properties and metadata, attached files, scripts, hidden layers, comments, form fields, the earlier versions it still holds, signatures and fonts, with a report to download. Clean PDF takes out the parts you choose, a batch of files at a time, and Redact PDF now removes the hidden data too when it saves, unless you untick it.
- Merge, split, extract and organize now keep a file’s layers, hidden ones still hidden, and its page labels, and say what they cannot keep: tags for screen readers, attached files outside Merge, and XFA forms. Merge keeps the files’ bookmarks and attached files, and can instead add one bookmark per file with the file’s own inside, or none.
- PDF to PDF/A now checks every file it makes with veraPDF, the reference validator, and says whether the file conforms, listing the rules it breaks when it does not.
- With Pro, the account page can keep every browser tool on this device, with the text recognition languages you choose, so the tools open and work with no connection at all. The copy is brought up to date when the site changes, and you can remove it at any time.
- Fill and sign now fills forms that also carry an XFA version, so that Adobe Acrobat shows the answers too, and says when a form is made only in XFA, which only Adobe’s readers can fill.
: A menu on phones and tablets, and every page checked against what the site does
- On a phone or a tablet, a menu button in the header opens every section: all tools, the guides, pricing, sign requests, the account and the language. The header shows the links themselves from a laptop’s width.
- Every tool page now says what Free and Pro each take: on Free, files up to 50 MB and 100 pages, 10 at a time; with Pro, up to 500 MB and 50 at a time. The Edit PDF page no longer mentions comments, which the editor does not add, and the Redact PDF page says which places a search of the saved file does not check.
- The privacy policy now lists everything an account holds, the browsers signed in, two-step sign-in, consents, the log of mail we send and our support’s records, with how long each is kept, and says that usage statistics are kept as single events for 90 days before only daily totals remain. The terms set an age of 16 for accounts, and the cookie and refund policies are more exact. The note shown before the first use of an AI tool now says when Anthropic may keep a text longer.
: Your name on the account, and signs that the site is at work
- The form that creates an account now asks for your name and, if you like, a phone number, which is checked against the numbering rules of its country. Both can be changed or removed under Your details on the account page, and the privacy policy says what they are used for.
- Every button that waits for our servers shows a spinner until the answer comes, a thin bar along the top of the window shows that the next page is on its way, and a tool that waits for one of our servers to start says so.
: Single sign-on for teams
- A team can let its people sign in with the company’s Google Workspace, Microsoft Entra ID or another OpenID Connect provider, after proving its email domain with a DNS record.
- People who sign in this way can join the team on the spot, while it has seats free, and a team can require single sign-on for its domain; its owner keeps the other ways in.
: Teams: Pro for everyone at work, on one invoice
- Set up a team on the new team page and choose its seats, from two, by the month or the year. Everyone who holds a seat has Pro, and the owner gets one invoice.
- Invite people by email: the link works for seven days, once, and only for the address it was sent to. Owners and admins look after the members and their roles; the owner adds or removes seats at any time, and Paddle charges or credits the difference.
- The team shares its AI credits, the month’s credits for every seat, and members can share sign templates and workflows with each other.
: Passkeys
- Passkeys: add one on the account page and sign in with your fingerprint, your face or your screen lock instead of the password, without typing your address. The passkey stays on your device or in your password manager and we keep only its public half. A passkey checked by your fingerprint, face or PIN also stands in for the authenticator code.
- The cookie policy and the privacy settings now list every item the site keeps in your browser, four of which were missing: the sign-up choices kept for the emailed link, the Pro kept for offline use, and the closed AI note and announcement. Signing out now also forgets the Pro kept for offline use at once.
: AI with Pro: summarize, chat and translate
- Summarize PDF says what a document says in a paragraph, or as up to twelve key points that each name their page, in any of 17 languages. A long document of up to 500 pages is read in parts, so nothing is cut off, and scanned pages are read with OCR in your browser first.
- Chat with PDF answers questions from the document itself, each fact with its page in square brackets, and says so when the document does not seem to answer. Your browser finds the passages that match each question, and only those are sent.
- Translate PDF translates each paragraph and puts it back where the original was, in its own font, size and color, into 17 languages, Hindi, Telugu, Tamil, Kannada, Malayalam, Bengali, Marathi, Gujarati and Punjabi among them; a translation that needs more room is set smaller rather than spilling over.
- All three run on Claude Haiku 5.5, Anthropic’s model, and use the AI credits that come with Pro: one credit for every two pages summarized, two a question, five a translated page. The cost shows before anything is sent, a request that fails gives its credits back, and only the text leaves your device; Anthropic does not train on it and deletes it within 30 days, and we keep counts and costs, never the text. The privacy policy has a new section on the AI tools.
: Five tools on our servers, and a stronger way for Compress and OCR
- HTML to PDF prints an HTML file the way a browser prints it, with its styles, tables, pictures and links, and Markdown to PDF typesets Markdown, tables, task lists and code included, with page numbers and bookmarks. Both run on our servers, in a browser that runs none of the file’s scripts and fetches nothing from the web.
- EPUB to PDF turns an e-book into a PDF on A4, Letter or A5, with the book’s own fonts and pictures, chapters on new pages, numbered pages and the table of contents as bookmarks; picture books keep their pages. Books locked with DRM are refused.
- PDF to SVG draws each page as an SVG that stays sharp at any size, and the PDF/A validator checks a file against PDF/A-1 to 4 or PDF/UA with veraPDF, the reference validator, and shows every rule the file breaks.
- Compress PDF and OCR PDF can, with Pro, do their work on our servers instead: Ghostscript for the strongest compression, at print, screen or smallest quality, and OCRmyPDF for scans of up to 2,000 pages, each page straightened first. Both still run in your browser by default, for free.
- Web page to PDF now sends every request the page makes through a gateway that looks each address up once and connects to that very address, so no site can point the printing browser at our own network; the page’s headings become bookmarks.
: Fourteen new tools, batches and saved workflows
- Page layout: Pages per sheet puts 2 to 16 pages on each sheet, on A4, Letter or the document’s own paper, with the text kept as text. Split pages in half cuts scanned spreads and two-up printouts down the middle or across, keeping a cover whole and reading right to left too. Resize PDF pages changes the paper size, scaling the content to fit, keeping it at its size or adding margins, and links keep working.
- For long and legal documents: Add header and footer writes text in six places, with the page number, the page count, the date or the file name, in any script and mirrored for books. Bates numbering gives every page of one or more files a prefix, padded digits and a suffix, continuing across the files in the order you set, with an index of the ranges. Bookmarks and contents edits and nests bookmarks, makes them from the document’s headings, and adds a contents page whose lines are links.
- For scans and print: Deskew PDF measures each page’s tilt from its lines of text and turns it back at full quality, with every angle yours to adjust. Remove blank pages finds empty pages while ignoring dust, scanner edges and punch holes, and shows you every page before it goes. Grayscale PDF turns text, drawings, gradients and photos to the gray of the same lightness in the file itself.
- New exports: PDF to text, in reading order without running headers and page numbers; PDF to HTML, one clean page of headings, paragraphs, lists, tables and pictures that is safe to open anywhere; PDF to TIFF, in color, gray or black and white from 150 to 600 dpi, as one multi-page file or one per page.
- Verify PDF signatures checks every digital signature in your browser against the Adobe Approved Trust List and the EU trusted lists: who signed, when, whether the file changed afterwards and how, with the signed version itself to download. It is free and never counts as a task.
- Sign PDF with a certificate, part of Pro, signs with your own .p12 or .pfx certificate inside your browser, in the PAdES form readers check, visibly or invisibly, as an ordinary or a certifying signature, with a trusted timestamp for which only the signature’s fingerprint leaves your device. Every signed file is checked by Verify PDF signatures before it is offered to you.
- PDF to Markdown and PDF to Excel read more of a document’s structure: tables ruled only between their rows come out as tables, numbered section titles set large or bold stay headings, and numbered lines run together in one paragraph become a list.
- Batches, with Pro: the tools made for one PDF, such as Compress, Rotate, Add page numbers, Add watermark, Grayscale, Deskew, Remove blank pages, OCR, the exports and Sign PDF with a certificate, take several files and do each in turn with the same settings. The results come in one ZIP, and a file that could not be done is named with its reason while the others are done all the same.
- Saved workflows, with Pro: set a tool up on its page and add it as a step to a workflow, up to 20 steps, then run the workflow on one PDF or many from the new Workflows page, in your browser. Only the steps are saved to the account, never a file, and never the password a Protect PDF step sets.
: Sign requests to the standard of the large services
- Signers adopt their signature and initials once: drawn with a mouse, a finger or a pen, typed in a handwriting style made for their script (Great Vibes, Dancing Script, Mr Dafoe and Herr Von Muellerhoff for Latin; Kalam and Amita for Devanagari; Lakki Reddy for Telugu; Kavivanar for Tamil; Chilanka for Malayalam), or uploaded as a picture. Typed initials follow from the name. Drawn signatures go into the PDF as vector strokes and typed ones as real text, so both print sharply.
- A signer at a computer can draw on their phone instead: the signature window shows a QR code that works once and for two minutes, and the drawing appears on the computer.
- The certificate of completion now gives the request ID on every page, the sender’s address and country, the counts of pages, signatures and initials, each signer’s signature and initials, how they made them, their device in plain words, their address with its country, the consent they accepted with its version, and every step to the second, in English and in the sender’s language. The audit report can also be downloaded on its own.
- The finished file is sealed with an RFC 3161 timestamp from a public timestamp authority, which receives only the file’s fingerprint. A new verify page checks any copy against the record of the finished file: the browser computes the fingerprint and only that is sent. The record, without names or content, is kept for seven years.
- Senders choose how dates are written and whether signers are reminded automatically, can send a reminder with a fresh link at most three times a day apart, see when an invitation reached the signer’s mail server, save the placed fields as a template, keep a finished request’s files for longer or delete them at once.
: Free, Pro and Team
- Every browser tool stays free, with no account and no watermark: ten tasks a day on files up to 50 MB and 100 pages, ten files at a time. A task is a file a tool makes for you; reading a PDF, checking its signatures and signing a document someone sent you never count, and a task that fails is not counted. The count is kept per browser, network and account together.
- Pro (€6.99 a month, €47.88 a year, or €3.99 for a seven-day pass, with prices of its own in nine more countries) removes the daily count, keeps the browser tools working offline, and adds the tools that run on our servers (200 tasks a day, files up to 500 MB), 1,000 AI credits a month and sending documents for signature (100 every 30 days). Signing what you are sent stays free.
- Credits for the server tools, welcome credits, referral rewards and credit packs are gone; credits now pay only for AI. Credits bought before today and not yet spent became days of Pro, five credits to a day, rounded up. Gift codes give days of Pro.
- A purchase that has not been used is refunded in full within 14 days. Deleting an account now cancels a running subscription at Paddle first, so nobody is charged for an account that no longer exists.
- The pricing and account pages, every tool page, and the terms of use, the privacy policy, the refund policy and the cookie policy describe the new plans.
: Plus, credit packs, a pricing page and privacy settings
- The server tools are now paid for in credits, by file size and tool: three are free every day without an account and ten with one, a file up to 50 MB costs one credit, and text recognition and PDF to Office count double. A job that fails costs nothing. Deleting and remaking an account no longer brings new welcome credits, and a fresh cookie no longer starts the day over.
- Plus (monthly or yearly) brings 500 credits a month, files up to 500 MB, results kept up to 30 days, sign requests without limit and the front of the queue; credit packs of 100, 300 and 1,000 are bought once and last a year. Payments go through Paddle, our merchant of record, on its own checkout page; every purchase can be refunded within 14 days. The pricing page shows each country its own currency.
- The account page shows the plan, the credits with their expiry, the subscription, the purchases, and a way to Paddle’s customer portal for invoices, the payment method and cancellation.
- Privacy settings: the first visit asks whether usage statistics may be collected, with accept, reject and settings given the same weight, and nothing is counted until then; “Privacy settings” at the foot of every page reopens the choice. A cookie policy lists every item the site stores, and a refund policy says how refunds work.
- The terms of use and the privacy policy describe Plus, credits, Paddle and the privacy settings, and name Ireland, where the company is established, as the governing law.
: Forms that other viewers can fill, previews, and evidence on every tool
- A form made with Create PDF form, or filled with Fill and sign, showed nothing when more text was typed into it in Chrome, Acrobat or Preview: the fields named a font embedded with only the letters already used, and the form did not list it where viewers look. The fields now name Helvetica, which every viewer carries, at a readable size; Indian-script defaults keep their embedded font.
- Add watermark, Add image watermark, Add page numbers, JPG to PDF, Split PDF, Compress image, Resize image and Convert image show the result before it is made: a live preview of the page or picture with the options you chose, redrawn as you change them.
- Every tool page, all 38, now says what the tool takes and gives, how it works, what it measured on the sample files, what it cannot do and what to do when it says something, in all nine languages for the last two; most pages show pictures of the tool at work. The seven server tools name the day and the servers their figures came from.
- Word, Excel and PowerPoint sample files and a deliberately damaged PDF open from “Try with a sample file” on the conversion and repair tools.
- The sign requests page shows what composing a request and signing one look like, what it cannot do, what to do when it says something, and the questions people ask, in nine languages; the account page says what to do when signing in fails.
- The home page lists every tool page, measured the same way, by category.
: PDF to Excel in the browser, and five more measured pages
- PDF to Excel now runs in your browser: tables drawn with lines are read from their grid, other tables from the gaps between columns, and figures become numbers you can sum. It had failed on the servers for every file, because LibreOffice cannot write a spreadsheet from a PDF.
- A fault in reading PDF text is fixed: a font whose character map listed more than 64 letters in one block lost its first few letters in Markdown, Excel, Compare and the editor. The letters were often the commonest ones.
- Merge PDF, PDF to Word, PDF to Excel, Create PDF form and Redact PDF show what they can and cannot do, with results measured on the sample files and pictures of the tool at work. The PDF to Word figures come from the production servers, and every figure names the build it was measured on.
- The sample documents are set in Noto Sans and Noto Serif, so the rupee sign prints; the standard PDF fonts had printed a question mark. The editor names such embedded fonts without the number some tools append to them.
- The sign requests page explains how signing works before asking you to sign in; the privacy policy and the terms describe accounts, passwords, Google sign-in, sign requests and support mail as they are.
- Guides carry the organisation’s byline, the description search engines show for the home page says where the work happens, and each page sends the browser only the words it needs.
: Evidence on the pages
- Every tool has a “Try with a sample file” button: documents we wrote for the site (a report, an invoice, a brochure, a rental agreement in two versions, a form, two scanned letters) that open the tool in a few seconds.
- Compress PDF, OCR PDF and Edit PDF show what they can and cannot do, with results measured on the sample files and pictures of the tool at work.
- Four guides, in English: compressing a PDF under 200 KB, why a PDF will not compress, making a scanned PDF searchable, and redacting a PDF permanently.
- The About page exists in all nine languages and says how the tools are tested, how to report a wrong result and what is kept for how long.
- The home page starts with the most common tasks and a box that finds a tool as you type.
- Each page shows the day its words last changed, and that day, not the deploy, is what search engines are told.
- The home page and the social images say where the work happens tool by tool, instead of claiming that everything runs on your device. OCR PDF lists the nine languages it really reads.
- The legal pages name one address each for search engines, and this page exists.
: First words sooner
- Pages paint their text sooner: the stylesheet is inlined, two scripts, a font and a picture no longer load before the page, and the logo is served as small WebP files.
- Brand, icon, font and language files are cached by the browser for a long time, so a second visit downloads almost nothing.
: Passwords
- An account is created with a password, chosen after the emailed link has proven the address, and opened with that password from then on.
- Ten wrong passwords in fifteen minutes lock the account for fifteen minutes, and a password known from a public breach is refused when it is chosen.
- The account page has one path to sign in and another to create an account, each saying what it does.
- After each deploy the search engines that take IndexNow are told which pages exist.
: A name and a face
- The site is called UV pdfTool everywhere, with its logo in the header, the footer, the browser tab and the pictures shown when a page is shared.
- An About page says who makes the site and what the name is and is not.
- Usage statistics that name nobody: counts of pages and tool runs, without cookies, third parties or personal data, read on an administration page.
: Guides on every page, signatures by email
- Every tool page carries a guide below the tool: what it is for, three steps, five questions with answers, and related tools, in all nine languages.
- Sign requests: send a PDF for signature by email; each signer gets a private link, signs in the browser, and everyone receives the signed copy with a certificate and an audit trail.
- Sign in with Google, beside the emailed link.
- New tools: Compare PDF, Scan to PDF, Create PDF form, and PDF to PDF/A on the servers.
: Live at uvpdftool.com
- The site goes live with the free browser tools and, on the servers, the Office conversions, PDF repair and web page to PDF.
- PDF to Markdown and an image watermark tool; each Office conversion has its own page.
- The editor gets an app-like shell: a ribbon, a side panel and a bottom bar.
- The privacy policy names the storage, database, worker and email providers.
: Before launch
- The PDF engine and the first eleven browser tools, then the editor that changes a document’s own text like a word processor, with pictures, tables, OCR, redaction, links and pages.
- Nine languages from the first day: English, Hindi, Telugu, Tamil, Kannada, Malayalam, Spanish, Portuguese and Indonesian.
- Crop, flatten, properties and compress tools.