1791113 - [foxfooding] misread text from image

Status: RESOLVED WONTFIX

(Core :: DOM: Core & HTML, defect)

Description

[Peter deHaan [:pdehaan]]

Attachment: issue-53.png

Firefox Version: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.15; rv:106.0) Gecko/20100101 Firefox/106.0
Window Size (inner width and height): 1024x658
GitHub Username: @greenbrown

Steps to Reproduce

1. enable about:config
   dom.text-recognition.enabled set to true
2. search for poster with different font in an image
3. copy text from image

Expected Behavior

misread some words

Actual Behavior

it should capture the word correctly

Attachment

Link to the original attachment

Comment 1

[Peter deHaan [:pdehaan]]

Not sure of the locale (or relevance), but that dialog might be Indonesian: https://pontoon.mozilla.org/id/all-projects/all-resources/?search=teks+disalin&status=translated&string=247283

Not sure what original site the reporter was looking at (Google Search results?); but I found the poster in a few places by searching "Burtons a delightful font duo". For example https://elements.envato.com/burtons-J8PW63 or https://www.fontspring.com/fonts/angga-mahardika/burtons
I'm still not sure what the second word in the header is.

Comment 2

[Peter deHaan [:pdehaan]]

Downloaded that font sample image and opened it in macOS Preview.app and was able to copy the text out and Preview seems to think the text is "Type Valies,", which I don't think is right either.

Uploaded image to https://www.imagetotext.info/ and it extracted:

Type Valves
PRESENT
BURTONS
"A Delightful Fort Dua"
TRD
BDG
MRK

Comment 3

[Greg Tatum [:gregtatum]]

I'm not sure there is anything actionable on this, as it's calling out the macOS text recognition APIs, which are fallible. The locale is used for the image based on the locale for the DOM element, so there can be some variation, and only a subset of locales are used with proper lexical support. Otherwise basic Latin scripts are used.

https://support.mozilla.org/en-US/kb/text-recognition