TG Studios

TG Imprint

Write a document. Get a PDF a screen reader can actually read: real headings, real lists, real links, and a description on every picture. Then read the accessibility check on your own file before you send it.

This is an open beta. It works, and I use it, but the wording and the keys are still moving while people test it and write in.

Free and MIT licensed. Version 0.1.0, Windows.

No sign-up and no key. The installer puts it in your own user folder, so it never asks for an administrator, and it updates itself from then on, downloading only what changed and always asking first.

What it does

Most programs can print a PDF. Almost none of them tag it, and an untagged PDF is a page of loose text: a screen reader cannot tell a heading from a paragraph, does not know a list is a list, and has nothing at all to say about a picture.

TG Imprint writes tagged PDFs and refuses to write any other kind. Headings are headings, lists are lists, links carry their text, tables have real header cells, and every picture carries a description or is marked as decoration. When the file is written, the app checks its own work and shows you what it found, in words, before you send it to anybody.

Every picture needs a description

You cannot insert a picture without describing it or ticking the box that says it is decoration. There is a list of every picture in the document, with the ones still needing a description marked, which is the fast way to finish a file somebody else started.

A second pair of eyes, on your own key

If you cannot see the picture you are describing, press Ctrl+D and Claude, ChatGPT or Gemini will write a first draft. You read it, change it, and decide whether to use it. It is a draft, not an answer: these services are confident and sometimes wrong, and the app says so.

It uses your own key. There is no account here and no service of mine in the middle. Paste a key from Anthropic, OpenAI or Google into Preferences and it is kept in Windows Credential Manager, where you can see it and remove it yourself.

Nothing leaves your machine unless you ask. When you do ask, the app tells you first what it is about to send, to whom, and how big it is, and waits for your yes. A whole document, a batch of pictures, and any picture that arrived inside a file somebody sent you all ask every single time.

Make somebody else's PDF accessible

Open a PDF and TG Imprint pulls it apart and rebuilds it as a document you can edit: headings, paragraphs joined across pages, lists, links, and every picture extracted and listed as needing a description. Fix it, export it, and you have a tagged PDF where there was not one.

It reads Word documents, Markdown and plain text the same way. A scanned PDF, which is photographs of pages, cannot be read, and it tells you so rather than opening an empty document.

Keyboard first, all the way down

Everything is on a key, the keys are the ones you already know from Word, and every one of them is listed on F1 and in the manual. Your screen reader announces the structure as you move through the document: heading level two, list of three items, link, graphic.

There are three settings for how much the app itself speaks, from everything down to nothing, and whatever you choose, it also writes to the status bar so nothing is ever lost.

What is not in it yet

Requirements

Windows 10 or 11, and the Microsoft Edge WebView2 runtime, which Windows 11 already has and Windows 10 installs in a minute. The PDF is made by Microsoft Edge, that runtime, or Google Chrome, whichever is already on your machine. Nothing else to install.

The source is open

MIT licensed, on GitHub at https://github.com/tonygeb23/tg-imprint. Read it, take it apart, or send a patch.

Tell me what breaks

It is a beta, and I would rather hear it than not: info@tonygebhard.me.