Repository navigation
Releases
Releases · DeenadayalanMoodley/ScanMaster-releases
ScanMaster 2.52.0
Fixed
- Explore Equation said the same symbol over and over with NVDA. The arrow keys went through the equation inside the document, where the cursor never moves; NVDA says the character at the cursor every time an arrow key is pressed, so it kept saying the first symbol (and what ScanMaster said was cut off by the next key). Explore Equation now opens the equation in its own window as a tree of its parts: the whole equation, then its parts (a fraction's numerator and denominator, a root's radicand, a power's base and exponent, the terms of a sum), then the parts inside those, down to the symbols. It is an ordinary Windows tree, which NVDA, JAWS and Narrator read natively: each part, how deep it is, and how many there are at that level. Up and Down move between parts, Right goes into a part, Left comes back out, Escape closes. Each part is said by MathCAT in the style and detail of Preferences > Speech > Mathematics, parts are named by their place ("Numerator: ...", "Exponent: 2"), and the braille of the part you are on is in a box below.
Removed
- The "Explore equations by" setting (Enhanced, Simple, Character): it chose how the key-by-key walk started, which the tree replaces. The braille of the part is no longer put on the status bar.
ScanMaster 2.51.0
Fixed
- Camera text came out as nonsense when Preferences > Scanning had page clean-ups on. Those corrections (deskew, straighten text lines, whiten background, reduce noise, remove blur, black and white, and the rest) were applied to every camera picture, after it had already been cut to the page and turned upright. Measured on a real page from a document camera, each one alone cost between 0% and 38% of the words and all of them together left 15%. Camera pictures now skip them. A new Apply the page clean-up from Preferences > Scanning checkbox (off to start with) brings them back; splitting a book spread into two pages still follows the preference.
- The guidance told you to move a page you had not moved. It was too strict about a page touching the edge of the picture, and said "Move the page to the right" a moment after "Page in view" when a hand was taken away. Now: a page that has lost only a sliver (under 4%, worked out from the page's shape and the sides that are not cut off) is in view; a page that has lost a little (under 10%) is told "Move the page a little to the left"; the guidance waits longer (1.6 seconds) before saying a page that was in view is not.
Added
- "Page in view, with very little room. Raise the camera a little if you can." The camera has no zoom, so the way to get a larger view is to raise it. When a page nearly fills the picture this is said once.
- Automatic mode waits for a page. It takes no picture of an empty mat or of a page that is mostly out of view, so a page that is placed well is taken once it is still.
ScanMaster 2.50.0
Added
- The camera tells you whether the page is in view. With "Guide me" on (the starting choice), Capture from Camera keeps the camera viewing and says, in a few words, whether the page is in view ("Page in view"), which way to move it when it runs off an edge ("Move the page to the left", "...away from you", "...to the left and towards you"), when it is too big for the view, and when there is no page ("I can't see a page"). It speaks only when what it sees has settled (a hand crossing the page says nothing), only when it changes, and repeats a problem every few seconds while it is still there. Left and right are the camera's own: the left of what it shows. If Turn pages upright is off it also says when the page is upside down ("Turn it around") or on its side ("Turn it a quarter turn clockwise").
- Turn pages upright. "Turn pages upright" (on to start with) finds which way up each page is and turns it before it is read: a page that is upside down or on its side was read as nonsense. It uses Tesseract's orientation detection and, when that is not sure (a page of little text), reads the page both ways up and keeps the way that reads better. Needs Tesseract, which is installed with ScanMaster.
- Pages are read in the background, and you are told when they are all read. The text of each picture is read as it is taken, one page after another while you turn the next page; the status line counts the pages waiting, and when the last is read ScanMaster says "All pages are ready to read" and you can start reading straight away. Closing the dialog does not stop pages still being read: they are added, and announced, when done. Run OCR on each picture is now on to start with.
Fixed
- Text from the camera was poor. Photos were taken at the camera's small default size (about 160 KB); they are now taken at its largest (3264 by 2448 on a document camera, in about a quarter of a second), with continuous autofocus. The text of a page is now made into paragraphs, headings and tables (lines broken only by the width of the page are joined into sentences, and words split at the end of a line are joined) instead of one paragraph per line.
- Page edges never cut away the page's writing now: an outline that has print outside it is not used, and the cut leaves a small margin. A page that runs off the top or left of the picture is found too.
- Sideways pages were turned the wrong way by Preferences > Scanning's "Correct page orientation": Tesseract counts angles anticlockwise and they were used as clockwise, so a page on its side was turned further the wrong way (upside-down pages were right).
ScanMaster 2.49.0
Added
- Choose how a camera picture is read. Capture from Camera now has the OCR choices of the Scan dialog: Run OCR on each picture, an OCR engine (Windows OCR, Tesseract, PaddleOCR, the Mistral OCR, Mathpix and Datalab cloud services, or AI vision, where an AI model reads the picture, using the provider set for OCR in Preferences > AI Tasks), the OCR language for the engines that use one, and AI assistance: proofread the text with AI, and, if a picture has no text, describe it with AI. An engine that is not ready (not installed, no key) says so when you choose it. All of it is remembered.
- Page edge detection. "Detect page edges" (on to start with) finds the page in each camera picture, against the desk or mat, and cuts the picture to it, straightening a page that is turned or photographed at an angle to an upright rectangle, before it is cleaned up and read. If no page outline is found, the whole picture is used and the status line says so.
ScanMaster 2.48.0
Added
- Automatic capture from a camera. Capture from Camera has an Automatic checkbox for scanning a book with a document camera: Scan Master watches the camera and, as soon as the picture has been still for a second, takes it. Turn the page, take your hand away, and the new page is taken when it settles; a page left lying there is taken once, and nothing is taken while a hand or a page is moving. Without the checkbox it is manual: press Take Picture (Alt+T). Both modes can be used together, and the choice is remembered.
- Camera shutter sound. Every picture taken from a camera makes a shutter sound (a new earcon event, "Camera shutter", in Preferences > Earcons; it is in the Standard and Essentials presets, and can be given another sound or turned off).
- Pictures go straight into the document. There is no Insert step any more: each picture is cleaned up as Preferences > Scanning says (a book spread can become two pages), read with OCR if "Run OCR on each picture" is ticked, and added as a page, with Scan Master saying "Page N added". A new Place pages list chooses where: after the current page, before it, at the beginning or at the end (the end is the starting choice). Cleaning up and OCR happen in the background, so the next page can be turned meanwhile, and pages are added in the order they were taken.
Changed
- The camera is now held open while the Capture from Camera dialog is showing, so a picture is taken at once instead of waiting for the camera to start each time; it is released when the dialog closes.
ScanMaster 2.47.0
Added
- Accessibility checker (Review > Check Accessibility, Ctrl+Shift+F8). Finds what would make a document hard to use with a screen reader, or fail an accessibility check, before it is saved or made into audio. It applies the rules of the checkers people know, to the format the document is going to become: Word (the Word Accessibility Checker and WCAG 2.2), PDF (PDF/UA), EPUB (EPUB Accessibility 1.1 and Ace), DAISY and NIMAS (the DAISY guidelines and Z39.86), web pages (WCAG 2.2) and audio and talking books. 37 rules: pictures with no or placeholder alternative text, a missing title, language or author, skipped, empty or missing headings and paragraphs that only look like headings, tables with no header row, an uneven grid, blank rows or columns or a table inside a table, links that say "click here" or are a bare address or point to a bookmark that is not there, footnotes with no text, low text contrast (WCAG ratios), very small text, spaces and empty paragraphs used for layout, lists typed by hand, long passages in capitals, justified text, underlined text that is not a link, unresolved comments, pages that are only a picture, and a note about equations for NIMAS. Each problem is an error, a warning or a tip (a missing title is an error for PDF, EPUB and DAISY, a warning for Word and web pages, as in those standards); choosing one shows how to fix it and which standard it comes from, and Go To takes the cursor there. Changing "Check for" checks the same document with another format's rules.
- It runs before you save or make audio. Saving a Word, PDF, EPUB, DAISY, NIMAS or web page file, or Save as Audio, now stops when the check finds errors and asks whether to look at them (Yes), go ahead anyway (No) or stop (Cancel); going ahead is remembered for those errors in that document. Warnings and tips never interrupt. Preferences > Advanced > "Check accessibility before saving or making audio" turns it off.
ScanMaster 2.46.0
Added
- Save As writes DAISY 2.02, DAISY 3 (NISO Z39.86) and NIMAS books, text with pictures. The Save As list has three new types: "DAISY 3 Book, text with pictures (NISO Z39.86)" (package, DTBook, navigation file and pictures), "NIMAS Book, text with pictures" (the same without the navigation file) and "DAISY 2.02 Book, text with pictures" (
ncc.html,content.htmlwith the book's styles and layout, a SMIL file for each part from one heading to the next, and the pictures). They have no audio: Speech > Save as Audio stays the way to make books that are read aloud in step with their text. The new DAISY 3 navigation file and the DAISY 2.02 files pass the official validators. A DAISY 2.02 book opened from itsncc.htmlcan now be saved back in place when it is a text book; one that has audio is refused with an explanation, as its audio would no longer match the text.
Fixed
- Nested lists are kept when a document is saved as a web page, EPUB or DAISY 2.02 book and opened again: a list inside a list item used to be flattened into its parent's text.
[2.45.0] - 2026-11-04
Added
- Open DAISY 3 (NISO Z39.86), NIMAS and DAISY 2.02 books, to edit them and describe their pictures. File > Open takes a book's package file (
.opf; "DAISY 3 and NIMAS Books" in the list of types) or a DAISY 2.02 book'sncc.html, and a ZIP holding a book. A DAISY 3 or NIMAS book opens from its DTBook with its structure as the book has it: heading levels, nested lists (bulleted, numbered, lettered or in Roman numerals), tables with their header rows and captions, pictures with their alt text, captions and producer notes, page numbers, notes (kept as real notes, with the markers pointing at them), sidebars, poems, definition lists, block quotes, bridgeheads, emphasis, sub- and superscripts, links, text in another language, and equations. The book's own parts have styles of their own in the Styles pane (Title, Author, Page Number, Sidebar, Producer Note, Poem Line, Definition Term and the rest), and each top-level section is a page. A DAISY 2.02 book opens for its text in reading order, with its pictures, headings, lists, tables and page numbers; its audio is not kept. - Saving a book. Save and Save As with "DAISY 3 and NIMAS Books (*.opf)" write the package, the DTBook content and the pictures (an
imagesfolder) beside the chosen file, and read back as the same book (the structure above, the formatting, notes, nesting and header rows were compared after a save and a re-open). A picture's alt text is saved as itsalt, and a long description (the Image Description paragraph after the picture, from Describe Image) as its producer note, where DAISY and NIMAS books keep descriptions; a picture nobody has described stays undescribed. Checked with the official DTBook 2005-3 validator (DAISY Pipeline 2) and the Z39.86 DTDs: the only thing reported is the fallback picture NIMAS asks for each equation, which is not made (each equation is saved with its text description). A DAISY 2.02 book cannot be saved back (its audio would no longer match); Save As another format, such as DAISY 3 and NIMAS. A DTBook has no place for underline, strike-through, colours, fonts or alignment, so those are not saved. - Books made with Save as Audio (DAISY 2.02 and DAISY 3) open the same way.
Removed
- The legacy
.smfformat. ScanMaster has no such format of its own to keep: its file type, the option to associate it, and its right-click menu entries are gone. (Nothing else changes:.smdxis the document format.)
[2.44.1] - 2026-11-03
Fixed
- The installer failed when installing for the current user only. The new right-click menu entries (and the .smdx/.smf file associations) were written to the machine-wide registry classes, which a per-user install may not write: "Access is denied" and the whole install was rolled back. They now go to the machine's classes for an all-users install and to the user's own for a per-user install.
[2.44.0] - 2026-11-03
Added
- Open files from the command line.
ScanMaster.exe [switches] file...opens the files when it starts. Switches:--auto(a PDF or picture opens with the automatic settings, without asking),--read(open the file and read it aloud),--audio(open the file and show Save as Audio) and--new-window(a window of its own). When ScanMaster is already running, the files open as new tabs in that window and the second program ends at once, so double-clicking or choosing several files fills tabs in one window; files that need a dialog or a long import open one after another. Files with the .smdx and .smf extensions now open when double-clicked (the association was set up, but the file name was ignored). - A ScanMaster submenu in Explorer's right-click menu. The installer (a new "Add a ScanMaster submenu to the right-click menu" choice, on by default; not in a portable copy) adds it to every file type ScanMaster opens: choose a file, then ScanMaster, then Open, Open with automatic settings (PDFs and pictures), Read aloud or Save as audio (documents and PDFs). On Windows 11 it is under Show more options (Shift+F10).
[2.43.0] - 2026-11-03
Added
- Open a picture: File > Open now takes PNG, JPEG, GIF, BMP, TIFF and WebP files. An Open Picture dialog asks what to do, like Import PDF does for a PDF, and says how big the picture is. Auto (the default) reads the text with Windows OCR and, if the picture has no text (a photograph, a drawing), has the AI describe it instead. You can also choose one engine to recognise the text (Windows OCR, Tesseract, PaddleOCR, Docling, Equations (local), or the cloud services Mistral OCR, Mathpix and Datalab), set the OCR language, and use the AI options (proofread the text, preserve the exact layout, also describe the picture), which start as the defaults on the AI page. Describe the picture only skips text recognition. The picture opens as a one-page document: the picture in the page panel, the text found in it, and its description where there is one (shown as plain lines, without Markdown symbols). It opens as an untitled document, so saving never overwrites the picture file. A picture with no text, with AI off or no provider that can see pictures, says so instead of opening blank.
- Pictures appear under "Pictures" (and in All Supported Files) in the Open dialog; they are not offered when saving.
[2.42.0] - 2026-11-02
Added
- Test connection now says what the model can do. After the model answers, the AI page's Test connection reports whether it chats with text, reasons (thinks before it answers), reads pictures, reads the text in pictures (OCR), calls tools (function calling), listens to audio, speaks, transcribes speech, makes embeddings or draws pictures. Each answer says how it is known: tested (it answered; its reply showed it thinking; it read a small test picture, which costs a few more tokens), the provider says so (the service's own details for that model, from Mistral, Ollama, OpenRouter, Gemini and Anthropic), or judging by its name (OpenAI and Hugging Face tell nothing about a model). What cannot be known is listed as not known. A model that is not for chatting (an embedding, speech or OCR-only model) is described rather than reported as a failed test, and a model its provider says cannot see is not sent the picture. Checked live against Mistral, OpenAI, Anthropic and OpenRouter.
[2.41.1] - 2026-11-02
Fixed
- Mistral: "top_p must be 1 when using greedy sampling" is fixed. Test connection, language detection and grammar checking ask for the most predictable answer (a temperature of 0), which Mistral refuses unless
top_pis also 1. ScanMaster now sendstop_p1 to Mistral, and answers the same refusal from any other service that passes requests on to Mistral by sending it again withtop_p1. - Mistral's reasoning models can be read. They answer with a list of parts, a "thinking" part and then the text; ScanMaster took that for an unknown reply. It now reads the text parts and leaves the thinking out. Checked live with the Mistral key in your settings: the request is accepted and the answer "OK" comes back.
- An unreadable reply now shows more of what the service sent (900 characters instead of 240) in the error.
[2.41.0] - 2026-11-02
Changed
- The AI and AI Tasks pages are one page: Preferences > AI. At the top are the provider and model most things use. Below them, choose a Job (OCR, Proofreading, Layout, Image descriptions, Grammar checking, AI Assistant, Dictionary, Speaker detection) and the page shows that job's own provider and model; Proofreading, Layout and Image descriptions also show their "by default" box, and Proofreading, Layout, Image descriptions and Grammar checking show their instructions (the text sent to the AI) with Reset to default. This replaces eight rows of lists with one set, so the page is shorter and a screen reader has far less to move through; the Account and limits section stays at the bottom. Everything you had set carries over. The old "Edit prompt for" list is gone, since the Job list chooses the instructions.
[2.40.0] - 2026-11-01
Changed
- One place for every API key: Tools > Preferences > API Keys. The keys of the AI providers (OpenAI, Anthropic, Google Gemini, Mistral, OpenRouter, Ollama Cloud, Hugging Face, and the address of an Ollama server), the Cloud OCR services (Mistral, Mathpix, Datalab) and Cloud TTS (ElevenLabs, Microsoft Azure Speech, Google Cloud Text-to-Speech, Deepgram, an OpenAI-compatible speech server; OpenAI's key is shared with AI) are all added, changed and removed on this page: a list of the services that have a key, with Add, Edit and Remove, saved as soon as you make a change. The AI, Cloud OCR and Cloud TTS pages no longer have key boxes; the Cloud TTS add/edit w...
ScanMaster 2.45.0
Added
- Open DAISY 3 (NISO Z39.86), NIMAS and DAISY 2.02 books, to edit them and describe their pictures. File > Open takes a book's package file (
.opf; "DAISY 3 and NIMAS Books" in the list of types) or a DAISY 2.02 book'sncc.html, and a ZIP holding a book. A DAISY 3 or NIMAS book opens from its DTBook with its structure as the book has it: heading levels, nested lists (bulleted, numbered, lettered or in Roman numerals), tables with their header rows and captions, pictures with their alt text, captions and producer notes, page numbers, notes (kept as real notes, with the markers pointing at them), sidebars, poems, definition lists, block quotes, bridgeheads, emphasis, sub- and superscripts, links, text in another language, and equations. The book's own parts have styles of their own in the Styles pane (Title, Author, Page Number, Sidebar, Producer Note, Poem Line, Definition Term and the rest), and each top-level section is a page. A DAISY 2.02 book opens for its text in reading order, with its pictures, headings, lists, tables and page numbers; its audio is not kept. - Saving a book. Save and Save As with "DAISY 3 and NIMAS Books (*.opf)" write the package, the DTBook content and the pictures (an
imagesfolder) beside the chosen file, and read back as the same book (the structure above, the formatting, notes, nesting and header rows were compared after a save and a re-open). A picture's alt text is saved as itsalt, and a long description (the Image Description paragraph after the picture, from Describe Image) as its producer note, where DAISY and NIMAS books keep descriptions; a picture nobody has described stays undescribed. Checked with the official DTBook 2005-3 validator (DAISY Pipeline 2) and the Z39.86 DTDs: the only thing reported is the fallback picture NIMAS asks for each equation, which is not made (each equation is saved with its text description). A DAISY 2.02 book cannot be saved back (its audio would no longer match); Save As another format, such as DAISY 3 and NIMAS. A DTBook has no place for underline, strike-through, colours, fonts or alignment, so those are not saved. - Books made with Save as Audio (DAISY 2.02 and DAISY 3) open the same way.
Removed
- The legacy
.smfformat. ScanMaster has no such format of its own to keep: its file type, the option to associate it, and its right-click menu entries are gone. (Nothing else changes:.smdxis the document format.)
[2.44.1] - 2026-11-03
Fixed
- The installer failed when installing for the current user only. The new right-click menu entries (and the .smdx/.smf file associations) were written to the machine-wide registry classes, which a per-user install may not write: "Access is denied" and the whole install was rolled back. They now go to the machine's classes for an all-users install and to the user's own for a per-user install.
[2.44.0] - 2026-11-03
Added
- Open files from the command line.
ScanMaster.exe [switches] file...opens the files when it starts. Switches:--auto(a PDF or picture opens with the automatic settings, without asking),--read(open the file and read it aloud),--audio(open the file and show Save as Audio) and--new-window(a window of its own). When ScanMaster is already running, the files open as new tabs in that window and the second program ends at once, so double-clicking or choosing several files fills tabs in one window; files that need a dialog or a long import open one after another. Files with the .smdx and .smf extensions now open when double-clicked (the association was set up, but the file name was ignored). - A ScanMaster submenu in Explorer's right-click menu. The installer (a new "Add a ScanMaster submenu to the right-click menu" choice, on by default; not in a portable copy) adds it to every file type ScanMaster opens: choose a file, then ScanMaster, then Open, Open with automatic settings (PDFs and pictures), Read aloud or Save as audio (documents and PDFs). On Windows 11 it is under Show more options (Shift+F10).
[2.43.0] - 2026-11-03
Added
- Open a picture: File > Open now takes PNG, JPEG, GIF, BMP, TIFF and WebP files. An Open Picture dialog asks what to do, like Import PDF does for a PDF, and says how big the picture is. Auto (the default) reads the text with Windows OCR and, if the picture has no text (a photograph, a drawing), has the AI describe it instead. You can also choose one engine to recognise the text (Windows OCR, Tesseract, PaddleOCR, Docling, Equations (local), or the cloud services Mistral OCR, Mathpix and Datalab), set the OCR language, and use the AI options (proofread the text, preserve the exact layout, also describe the picture), which start as the defaults on the AI page. Describe the picture only skips text recognition. The picture opens as a one-page document: the picture in the page panel, the text found in it, and its description where there is one (shown as plain lines, without Markdown symbols). It opens as an untitled document, so saving never overwrites the picture file. A picture with no text, with AI off or no provider that can see pictures, says so instead of opening blank.
- Pictures appear under "Pictures" (and in All Supported Files) in the Open dialog; they are not offered when saving.
[2.42.0] - 2026-11-02
Added
- Test connection now says what the model can do. After the model answers, the AI page's Test connection reports whether it chats with text, reasons (thinks before it answers), reads pictures, reads the text in pictures (OCR), calls tools (function calling), listens to audio, speaks, transcribes speech, makes embeddings or draws pictures. Each answer says how it is known: tested (it answered; its reply showed it thinking; it read a small test picture, which costs a few more tokens), the provider says so (the service's own details for that model, from Mistral, Ollama, OpenRouter, Gemini and Anthropic), or judging by its name (OpenAI and Hugging Face tell nothing about a model). What cannot be known is listed as not known. A model that is not for chatting (an embedding, speech or OCR-only model) is described rather than reported as a failed test, and a model its provider says cannot see is not sent the picture. Checked live against Mistral, OpenAI, Anthropic and OpenRouter.
[2.41.1] - 2026-11-02
Fixed
- Mistral: "top_p must be 1 when using greedy sampling" is fixed. Test connection, language detection and grammar checking ask for the most predictable answer (a temperature of 0), which Mistral refuses unless
top_pis also 1. ScanMaster now sendstop_p1 to Mistral, and answers the same refusal from any other service that passes requests on to Mistral by sending it again withtop_p1. - Mistral's reasoning models can be read. They answer with a list of parts, a "thinking" part and then the text; ScanMaster took that for an unknown reply. It now reads the text parts and leaves the thinking out. Checked live with the Mistral key in your settings: the request is accepted and the answer "OK" comes back.
- An unreadable reply now shows more of what the service sent (900 characters instead of 240) in the error.
[2.41.0] - 2026-11-02
Changed
- The AI and AI Tasks pages are one page: Preferences > AI. At the top are the provider and model most things use. Below them, choose a Job (OCR, Proofreading, Layout, Image descriptions, Grammar checking, AI Assistant, Dictionary, Speaker detection) and the page shows that job's own provider and model; Proofreading, Layout and Image descriptions also show their "by default" box, and Proofreading, Layout, Image descriptions and Grammar checking show their instructions (the text sent to the AI) with Reset to default. This replaces eight rows of lists with one set, so the page is shorter and a screen reader has far less to move through; the Account and limits section stays at the bottom. Everything you had set carries over. The old "Edit prompt for" list is gone, since the Job list chooses the instructions.
[2.40.0] - 2026-11-01
Changed
- One place for every API key: Tools > Preferences > API Keys. The keys of the AI providers (OpenAI, Anthropic, Google Gemini, Mistral, OpenRouter, Ollama Cloud, Hugging Face, and the address of an Ollama server), the Cloud OCR services (Mistral, Mathpix, Datalab) and Cloud TTS (ElevenLabs, Microsoft Azure Speech, Google Cloud Text-to-Speech, Deepgram, an OpenAI-compatible speech server; OpenAI's key is shared with AI) are all added, changed and removed on this page: a list of the services that have a key, with Add, Edit and Remove, saved as soon as you make a change. The AI, Cloud OCR and Cloud TTS pages no longer have key boxes; the Cloud TTS add/edit window says whether a key is saved. Mistral has one key for AI and for Cloud OCR.
- Keys are hidden. A key is typed into a password box, so it is not shown on the screen and NVDA and other screen readers do not read it out (before, the AI and Cloud OCR key boxes were ordinary text boxes that a screen reader read aloud and anyone looking could see). The keys were already stored encrypted.
- AI Tasks and the AI page offer only providers that have a key. Each job's Provider list shows Default and the providers (and, for OCR, the document services) with a key on the API Keys page; the AI page's Provider list is the same. A provider whose key was removed after you chose it for a job stays in that job's list marked "(no API key)", so nothing changes without you. When you add or remove a key, these lists follow at once.
- Keys you had already entered (AI, Cloud OCR, and the key inside each Cloud TTS provider) are moved to the new page the first time you start this version. An Ollama address that was only the default saved by pressing OK, with Ollama never chosen, is dropped, so Ollama is not listed u...
ScanMaster 2.44.1
Fixes the installer for current-user installs. See the changelog.
ScanMaster 2.44.0
Open files from the command line; a ScanMaster submenu in Explorer's right-click menu. See the changelog.