Visual Geez is two encodings, not one

Drop the file here, or paste the broken text. The converter marks both Visual Geez readings, and each one shows what it makes of your document, so you compare them by eye. If the two produce identical text they collapse into a single row — which is the tool telling you the choice doesn’t matter for this document.

The two encodings share 194 byte sequences. They agree on 182 of them and disagree on the 12 below. If your document contains none of these, both readings hand you the same text. If it contains one, you will see the difference immediately.

That table is also how you tell them apart. What is a letter on one side is punctuation on the other, so a document read as the wrong one either drops letters into hyphens and slashes or turns its punctuation into letters. Either way you can see it at a glance.

Identifying an encoding is done by counting how many bytes came out meaning something. That cannot separate the two Visual Geezes: both know the same bytes and differ only in what they say those bytes mean. Both will report reading your document completely. One produces words and the other doesn’t, and only a person who reads Amharic can say which.

Open the document, select the broken letters and look at the font name. If it says Visual Geez, you are on the right page. It won’t settle it, though. The two encodings have their own fonts, those fonts draw the letters much the same way, and documents pick up different font names as they move between machines. The name tells you where to look, not which one you have.

Each encoding is built on its own font, and every mapping was drawn in that font and checked against the character it claims to be.

One Visual Geez mapping did not come from the picture comparison at all. The bytes are 75. In one document an author spelled the same word four times one way and once with these bytes, and the character the first spelling gives was already known. So ጭ is what was published, on the evidence of what somebody actually wrote.