Skip to content

Usugami 1.5.0

Choose a tag to compare

View all tags
@github-actions github-actions released this 04 Nov 16:44
· 212 commits to main since this release
v1.5.0
97dfaeb

Added

  • Three editions. Usugami now comes as Usugami (Usugami.apk, recommended), Usugami Lite
    (Usugami-lite.apk) and Usugami OCR (Usugami-ocr.apk). They share a package name and signing
    key, so you can install any one over another and keep your files and settings. Settings → About
    shows which one you have.
    • Usugami works as before, with text recognition through Google Play services.
    • Usugami Lite has every tool except text recognition, and no Google code at all.
    • Usugami OCR has open-source text recognition built in: PaddleOCR's PP-OCRv5 models, run on
      your phone by Tencent's ncnn. It reads English, French, Spanish and Hindi (Hindi mode also
      reads English words), fully offline and with no Google code. It is about 14 MB.
  • Scan with the camera. Take photo in Images to PDF opens your phone's own camera app, page
    after page (Take another, then Done); Usugami still asks for no permissions. Drag the four corners
    onto the page's corners and it comes out straight, as if scanned; Usugami finds the page by itself
    when it stands out from the table, a magnifier shows what is under your finger, and arrow buttons
    and screen-reader actions move a corner a step at a time. Every page has a Crop button, for
    gallery photos, files and shared images too, and can be cropped again from scratch.
  • Looks for scanned pages: Original, Enhanced (white paper, no lamp color or shadows,
    darker print), Grayscale or Black & white, previewed as they will be in the PDF, for one
    page or all. Black-and-white pages are stored one bit a pixel (CCITT Group 4, as scanners do),
    so a page of print takes about 10 KB instead of 50 to 350 KB as a photo. Text recognition reads
    cropped pages as cropped.
  • Batch mode. Compress PDF, Add page numbers, Add watermark and Flatten PDF work on many PDFs at
    once, and Word, PowerPoint and EPUB to PDF convert many files at once, one PDF each. Pick several,
    add or remove files as you go, and name each new PDF. Compress PDF sets the level for all PDFs
    in one go or gives a PDF a level of its own. Files are made one after another with "PDF 2 of 5"
    progress, Cancel stops them all, and a file that can't be made is named while the others are
    still made. Home's search finds these tools with "batch", "many" or "several".
  • When a tool makes several PDFs (batch mode, or Split PDF in parts), each PDF on the result screen
    has its own Open, Share and Save buttons and a "Saved" mark; Save to Downloads and
    Save to folder… still save them all, each as its own file, never twice.
  • Share PDFs, Word, PowerPoint and EPUB files to Usugami from another app, as well as images.
    PDFs open the new Choose a tool screen (every PDF tool for one PDF; Merge PDF and the batch
    tools for several), and documents open their converter with the files added. Shared files are
    copied in at once; files shared while a tool has unsaved work ask first, and ones shared while a
    PDF is being made wait for it.

Changed

  • Usugami OCR replaces Usugami Offline, which used Google's closed-source ML Kit.
    Usugami-offline.apk is no longer published: install Usugami-ocr.apk (or Usugami-lite.apk) over
    it, and your files and settings stay. If Obtainium follows Usugami-offline.apk for you, add
    Usugami OCR or Usugami Lite instead.
  • Text recognition reads your original photos, not the compressed copies inside the PDF, so
    small print is found even at Extreme compression. In tests at Extreme, Usugami OCR's errors fell
    from about 29% to under 4%, and Usugami's own recognition got better too. Long scrolling
    screenshots are read at a readable size instead of being shrunk to a sliver first.
  • Usugami OCR reads pages photographed upside down or sideways the right way up, and keeps lines
    in order on slanted photos and in columns.
  • At large font sizes and in long translations, titles, buttons, choices and labels shrink, wrap or
    move to the next line instead of being cut off.
  • The open-source notices no longer say they apply to Usugami 1.2.0.

Fixed

  • Split PDF: each part now holds only its own pages. A part cut from pages with links,
    comments or form fields could carry the rest of the PDF hidden inside it (other pages' text,
    pictures and form values) and be as big as the original. Splitting into many parts also no
    longer gets slower with every part.
  • Delete pages, Reorganize PDF and Merge PDF: pictures and tags (alternate text) of the pages
    you remove no longer stay hidden in the file when pages share them, and every page keeps its page
    label (such as "iv" or "245") instead of the labels shifting.
  • Split PDF messages say what to type for comma-separated custom ranges and backwards ranges
    (8-3); tapping a preview no longer re-sorts the pages typed; Fixed chunks shows an example and
    opens the number keyboard.
  • Merge PDF says a second PDF is needed when only one is added, follows the file order in page
    mode, and no longer asks to discard changes when nothing changed.
  • Compress PDF now shrinks very large JPEG images (600-dpi scans, 50-megapixel photos) instead
    of saying the PDF is already compact, uses less memory for the others, and keeps the original
    when it would save less than 1% (no more "0% smaller").
  • Password-protected PDFs show their pages in Split's picker, Merge's page mode, Reorganize and
    the PDF's card, instead of a placeholder icon.
  • PDFs with a few bytes before their start (a byte order mark or a blank line, as some websites
    send) now open.
  • Redact PDF is more thorough: a page without boxes that still holds, out of sight, part of what
    a box hides now becomes a picture too; a picture another page shows cropped no longer keeps what a
    box hides; attached files, a digital signature's signer details and the names of links to
    redacted pages are removed. Find text also searches comments, and finds D’Souza for D'Souza,
    Straße for Strasse, Wałęsa for Walesa and numbers written with special hyphens or minus signs.
    Form values Usugami couldn't draw are shown and found instead of looking empty.
  • Flatten PDF no longer blanks a filled-in field it can't draw (such as ₹ or Hindi in a form
    made for English): the field keeps its look or stays fillable, and Usugami tells you. The tools'
    previews show such fields as they really look.
  • Edit PDF and Sign PDF: what you add over a form field, comment or stamp now covers it, as the
    preview shows. WebP photos are compressed like other photos (often several times smaller), Ultra
    HDR photos no longer carry their hidden HDR data into the PDF, tapping an addition no longer
    clears Redo, a pen stroke drawn as the page loads is no longer lost, and quick presses of an arrow
    button each move one step.
  • Word to PDF: letterheads, logos and watermarks in headers and footers no longer push the text
    off the page; rotated, flipped, grouped and hidden pictures and text boxes come out as in Word;
    page numbers follow their format (i, ii, A, B) and odd/even-page sections; Greek letters, maths
    symbols and check boxes from the Symbol and Wingdings fonts show as themselves; lists linked to a
    list style keep their numbers; automatic text color turns white on dark shading; right-to-left
    paragraphs, lists and tables are mirrored; phonetic guides keep their base text; and a document
    whose whole body is kept in another format now says so instead of giving a blank PDF.
  • PowerPoint to PDF: elbow and curved connectors are drawn as lines with arrowheads; ink,
    equations, 3-D models, Slide Zooms and embedded objects show as PowerPoint's own pictures of them;
    slide numbers count hidden slides; dark backgrounds and layouts keep their colors; flipped groups
    are mirrored; empty placeholders are left out; table borders follow the table style; right-to-left
    paragraphs are aligned from the right; Symbol and Wingdings characters, link colors and text
    anchored at the bottom or middle of a box come out as in PowerPoint.
  • EPUB to PDF: a picture, line break or rule hidden by a book's style sheet no longer cuts the
    rest of its section out of the PDF; tables keep the space above them; inline pictures fit the
    line; very long paragraphs convert much faster; formulas no longer print their TeX source;
    bookmarks into tables and old-style anchors lead to the right page; fixed-layout pages stay on one
    page; and running out of storage gives an error instead of a PDF with missing pages.
  • E-books, Word documents and presentations whose internal file names start with "/" now open on
    Android 14 and newer, and very deeply nested Word and PowerPoint files convert instead of
    crashing.
    Pictures fitted with Crop > Fit keep their shape, and fonts like Century Gothic and Monotype
    Corsiva get a similar font.
  • Images to PDF: with Fit page size, long scrolling screenshots keep their width instead of
    becoming unreadable slivers, and images added while others are still being added go after them
    instead of in between.
  • An unsaved result, even one with hundreds of PDFs, comes back after Android stops Usugami in the
    background, and going to the background with thousands of images or pages no longer crashes.
  • Back, Start over and Create wait while PDFs are being saved, sharing images during a save no
    longer deletes the PDFs being saved, and images shared while you kept a tool's PDF show once you
    leave the tool.
  • When the phone's storage is full, adding images, PDFs, documents or pictures says so instead of
    calling the file damaged, and phones without a file picker get a message instead of a crash.
  • Screen readers hear when a page has been turned in Images to PDF and Merge PDF.
  • Fast and double taps do each thing once. A double tap on Save as, Share, Open or a Pick or
    Add button opens one screen instead of two (whose second pick was lost); Save to Downloads tapped
    twice no longer saves a second copy; Reverse order, Edit PDF's text Done and Shape menu, and
    Reorganize's Reset act once; and a quick second tap on a turn button turns the page again
    instead of being lost.
  • Redact PDF: tapping Delete box twice quickly no longer also deletes another box, which could
    be on another page and leave its area unredacted.
  • Two pages tapped together in Delete pages and Split PDF are both marked; a page turned and then
    quickly moved or removed is moved or removed; and Add pages > Blank page asked twice no longer
    adds extra pages later.
  • Edit PDF and Sign PDF: tapping an item only chooses it (a slightly moving finger no longer
    nudges it or clears Redo), Save shows it is working right away so edits made meanwhile can't be
    left out, and Clear on the signature pad also clears a stroke still being drawn.
  • A ZIP file being saved is no longer cancelled (and deleted) by Back, Start over or images shared
    meanwhile, its progress shows over any screen, and a file chosen in Save as while another save
    runs gets its PDF instead of staying empty.
  • Opening a tool again right after leaving it starts it fresh, a PDF picked after leaving its tool
    no longer lands in another tool, the same message no longer repeats for seconds, and choosing
    languages quickly applies your last choice.
  • Text recognition: a very small or very thin image no longer leaves every PDF made with it
    without searchable text.
  • Redact PDF on Android 17: redacted pages, and the Redact and Edit PDF previews, are drawn
    exactly as big as the page, so a box placed near the foot of a page no longer leaves a sliver of
    what it hides showing.

Which file should I download?

File Size Pick it if… Text recognition (OCR)
Usugami.apk (recommended) 3.1 MB You're not sure. Works on most phones. Runs through Google Play services. The first time you use it, Play services downloads the text-recognition model (one for English, French and Spanish, one for Hindi), which needs internet then.
Usugami-lite.apk 2.8 MB You don't need text recognition, and want the smallest app with no Google code at all. None.
Usugami-ocr.apk 14.4 MB Your phone has no Google Play services, or you want text recognition with no network use at all. ARM phones (virtually all phones). Open source and built into the app (PaddleOCR's models, run by ncnn). Works without internet, ever.

Every other tool works offline in all three, and Usugami itself has no internet permission. In
Usugami.apk, text recognition uses Google's ML Kit, which runs on your phone but is closed source;
Usugami Lite and Usugami OCR contain no Google code. The rest of Usugami is GPL-3.0-or-later. All
three files are signed with the same key: install one over another to switch editions.

Had Usugami Offline? It is now Usugami OCR: install Usugami-ocr.apk over it, and your files and
settings are kept. In Obtainium, change its APK filter to ^Usugami-ocr\.apk$.

Needs Android 8.0 or newer. New to installing APKs? Follow the
step-by-step guide.

Verify your download (optional)

  • Package name: io.github.nath_goa.usugami
  • Signing certificate SHA-256: 6F:4F:23:AD:E0:3E:63:6A:7C:C3:68:16:89:44:D0:CC:FC:81:32:EF:34:C5:0B:C5:A4:6B:0C:A4:CB:BD:F4:ED

On a phone with Android 9 or newer, paste those two lines into
AppVerifier. On a computer:

sha256sum -c --ignore-missing SHA256SUMS.txt
apksigner verify --print-certs Usugami.apk
gh attestation verify Usugami.apk --repo Nath-Goa/usugami

SHA256SUMS.txt:

054432fe1838b971e18a720549420ae34fa2bcb1463fd275f0a25521dcc5d602  Usugami.apk
794794cb13b41ad8ff8ac45f4010dbcc591387434fa52b28ef092684e1168941  Usugami-lite.apk
a43f33fb746d0f30bb530661e81a998e88ab171af7c47c3a9c03d12df124438b  Usugami-ocr.apk

The mapping zip file is only for developers: it holds the R8 mapping files that turn stack
traces in crash reports back into readable code. You don't need it to install Usugami.

Assets 7
Loading