What FileConcat reads

Anything that is already text, such as source code, markdown, CSV, JSON or logs, goes into the bundle as it is. The formats below hold their text inside a container, and FileConcat reads it out in your browser. Your files are not uploaded to us.

A file is recognised by its first bytes, not its name, so a renamed or extensionless document is still read. Most formats are read as soon as you drop them. Images and recordings wait until you ask, because recognition and transcription take seconds to minutes. A recording also needs a speech model downloaded once, which the privacy page describes.

A document's own title, when its properties hold one, comes first as [Title: ...]. The author's name and the other properties are left out.

A picture's alternative text, the words its author wrote for a reader who cannot see it, comes through as [Image: ...] where the picture sits (in a saved web page, as a Markdown image), and a table's caption and description as [Table: ...] above the table. In a PDF the figures' text follows the page it is on. A picture with no alternative text, or only a default name such as "Picture 3", is still left out.

A link comes through as [text](address) where it sits, in documents, slides, PDFs and email. In a spreadsheet the cells come first, then each linked cell as [Link, B2: address]. A link to another place in the same file, or one whose text already is its address, is written as plain text.

Each entry also says what is left out, so you know what the model will not see before you paste. How well each one reads against other tools, at the latest version measured, is on the benchmarks page. The studies behind it are the extraction benchmark and, for how much of each format comes through comment by comment and chart by chart against Apache Tika and markitdown, markitdown vs Tika.

Documents

  • PDFread on drop

    .pdf

    Comes through
    The text, page by page under a page heading, which also gives the number the document prints on the page when that differs (iv in a preface). Text on a layer a viewer hides comes through, with a line saying so. Review comments come through as [Comment, author: text], after the words a highlight covers or after the page. A form's filled-in fields come through as [Form field, name: value], a check box as [x] or [ ]. Attached files are named, and a text file's contents come through.
    Left out
    Pictures and charts are not described. Text set at an angle, such as a diagonal DRAFT watermark, is left out. A password-protected PDF is not opened.

    PDF to text

  • Scanned PDF pagesread on drop

    .pdf

    Comes through
    Pages with no text of their own are read by text recognition and marked in the bundle. Up to three such documents and 8 MB per drop start on their own; past that it asks first.
    Left out
    A table comes through as loose lines. A faint or crooked scan reads less well.

    PDF to text

  • Wordread on drop

    .docx .docm .dotx .dotm .doc

    Comes through
    Text with headings, tables, footnotes, page headers and footers, and link addresses. A list item keeps the number or bullet Word shows for it. A chart comes through as a table of its data. In a .docx, a form check box comes through as [x] or [ ] and a drop-down as its chosen entry. Tracked deletions and hidden text are left out.
    Left out
    Pictures are not described. A chart in a Word 97-2003 file, or in a header or footer, is not read. A Word 97-2003 list numbered through a paragraph style loses its numbers.

    Word to text

  • OpenDocument text and RTFread on drop

    .odt .rtf

    Comes through
    The text, with its page headers and footers. An OpenDocument list item keeps its number or bullet, and an OpenDocument chart comes through as a table of its data. An OpenDocument form field comes through as its value and a check box as [x] or [ ]. Hidden text is left out.
    Left out
    Pictures and the charts in an RTF file are not described.

    RTF to textODT to text

Spreadsheets

  • Excelread on drop

    .xlsx .xlsm .xlsb .xltx .xltm .xls

    Comes through
    Every sheet under its name, rows as comma-separated cells, each cell as Excel shows it. Charts, text boxes, the print header and footer, a cell's input message and allowed values, and named ranges follow a sheet's rows, a chart as a table of its data. A hidden sheet is read and marked (hidden), hidden rows and columns are read and, except in .xlsb, named in a line after the rows, and a very hidden sheet, which only a macro can show, is left out.
    Left out
    Formulas (their results come through), cell colours and conditional formatting. The charts in an .xls or .xlsb file.

    Excel to text

  • OpenDocument spreadsheetread on drop

    .ods

    Comes through
    Every sheet's cells as text, then its validation messages and allowed values, its print header and footer, and its charts as tables of their data. A hidden sheet is read and marked (hidden), and hidden rows and columns are read and named in a line after the rows.
    Left out
    Formulas (their results come through) and cell colours.

    ODS to text

Slides

  • PowerPointread on drop

    .pptx .pptm .potx .potm .ppsx .ppsm .ppt

    Comes through
    The text of each slide, slide by slide, and a chart as a table of its data. Text the slide master prints on every slide comes once, at the top. A hidden slide is read and marked (hidden). In a .pptx, a section's name comes before its first slide, a custom show is listed with its slides, and a table cell merged into another is left empty.
    Left out
    A chart or diagram that exists only as a picture, and the charts and slide master text in a PowerPoint 97-2003 file.

    PowerPoint to text

  • OpenDocument presentationread on drop

    .odp

    Comes through
    The text of each slide, and a chart as a table of its data. Text the master page prints on every slide comes once, at the top. A slide the author named carries its name. A hidden slide is read and marked (hidden).
    Left out
    A chart or diagram that exists only as a picture.

    ODP to text

Books

  • EPUBread on drop

    .epub

    Comes through
    The chapters as plain text, in reading order. A part the book keeps outside the reading order is marked where it stands.
    Left out
    Pictures are not described.

    EPUB to text

  • Kindleread on drop

    .mobi .azw3

    Comes through
    The book's text, read the same way an EPUB is.
    Left out
    Pictures are not described. A book protected by DRM is not opened.

    EPUB to text

Email

  • Saved email and Outlook messagesread on drop

    .eml .msg

    Comes through
    From, To, Cc, Date and Subject, and the sender or reply address when it is someone else, then the decoded body, without text the message hides.
    Left out
    Attachments are named, not read. Images the body shows and signature files are not listed.

    Email to text

Notebooks

  • Jupyter notebookread on drop

    .ipynb

    Comes through
    Markdown: the prose, the code in fences and the text output. A raw cell for another renderer, such as LaTeX, is fenced under its format.
    Left out
    Embedded images are dropped and counted. An image with a text description, such as a plot, is named by it where it stood.

Subtitles

  • Caption filesread on drop

    .srt .vtt

    Comes through
    The transcript, without cue numbers or timestamps.
    Left out
    Who is speaking, unless the captions say so.

Audio and video

  • Video subtitle trackread on drop

    .mp4 .m4v .mov .mkv .webm

    Comes through
    A video's own text subtitle track, as a transcript.
    Left out
    A video with no subtitle track gives nothing here; transcribe it instead.
  • Recordingsread when you ask

    .mp3 .m4a .aac .flac .ogg .wav .mp4 .mov .mkv .webm

    Comes through
    A transcript made in the browser when you ask, plus a video's on-screen text. The speech model is downloaded once.
    Left out
    Speaker names. A noisy recording transcribes less well.

Web pages

  • Page saved from a browserread on drop

    .html .htm

    Comes through
    The article as markdown, a formula as its TeX source, and a form field as the value it shows.
    Left out
    Menus, ads, text the page hides and everything else outside the article. HTML source code stays code.

    Web page to text

Archives

  • Archivesread on drop

    .zip .tar .gz .tgz .7z .rar .bz2 .xz

    Comes through
    Unpacked in the browser; every file inside is read as if dropped on its own.
    Left out
    A password-protected or damaged archive is not opened.

Images

  • Imagesread when you ask

    .png .jpg .jpeg .gif .tif .tiff .webp

    Comes through
    Writing in the image, read by text recognition when you ask.
    Left out
    A photo with no writing gives nothing. Layout is not kept.

    Image to text

The command-line version reads fewer of these. This page describes FileConcat in your browser.