"Sorry, we can't translate anything in this photo. Pick a different photo or switch language."

I have not had prior experience to where the experiment, as seen on Google's own Art & Culture collection, where this experiment actually works. No matter which image I choose or take for the experiment to implement, it immediately spits out the dreaded "Sorry, we can't translate anything in this photo. Pick a different photo or switch language" error message– not even 5 seconds to convince me some AI or any API made use of the image or crunched any numbers. From my own research, this seems to be an issue with Google's backend implementation, acting as sort as the middle man to carry over the data to the AI in order for it to be interpreted, then translated, accordingly.

I know this project hasn't been visited lately, even with the occasional YouTube shorts showcasing it, but it'd be nice if it were at least able to take any images and process them, even if it cannot genuinely recognize what I'm trying to provide it.

The report names no files, tests, or entry points. Start by reproducing the photo-translation flow with different images and language settings, then determine whether the failure is in the app or the Google backend. Done means successful image processing or a documented, actionable failure.

Written by the indexing model from the issue text.