Google Lens and Apple Translate both have a pretty handy feature: translating text directly from your camera picture. I found “Translate you” on Github which has image OCR + translation but it’s not really useful. Are there any other alternatives?
If not I’ll probably use Google Lens when the App Scope feature drops in GrapheneOS.
I’ve been using DeepL. Isn’t opensource but seems a bit more privacy respectful than the others.
It isn’t.
“When using our translation service, please only enter texts that you wish to transfer to our servers. The transmission of these texts is necessary in order for us to provide the translation and offer you our service. We process your texts, the documents you upload and their translations for a limited period of time to train and improve our neural networks and translation algorithms”
I won’t dive on this too much because we will enter in a subject of personal opinion about a company practice. Personally, comparing to Google and Apple I still feel that they are harvesting and using the data for one case (train the tool).
I don’t have additional internal insight or access to audit reports to take a better informed decision and I wish there were other options.
At the moment, again if I have to choose between those 3, I will still select the DeepL for now until I learn or feel different about it. I strong advise to not use sensible information in this tool. If you are traveling to a country that you don’t know the language and need to translate a restaurant menu, I don’t see a problem.
While that is a real privacy risk in some contexts, that some people may not be comfortable with, the part of the privacy policy you quoted is not on its own unreasonable or suspicious in the context of a translation service.
They are just disclosing that your usage of the service is used to improve the service.
Translation services–including the privacy friendly ones–typically use machine learning or neural networks under the hood. And the quality of those services depend on the quality and breadth of the training data. In my eyes, a translation service has a legitimate reason to train on real users and natural language in the wild, to improve the service in ways that meaningfully benefit users. (an opt out would be nice though)
Also note that the bit you quoted applies to the free service, the paid service does not use your data to train the model:
After complete performance of the contractually agreed services all submitted texts or documents and their translations will be deleted. When using DeepL Pro, your texts will not be used to improve the quality of our services.
I’m not here to argue its wrong to be cautious or concerned about this, and I can see many contexts in which one might decide they aren’t comfortable with their translations being used to refine the model. But it isn’t in and of itself evidence of anything malicious or exploitative, and it doesn’t make DeepL ‘no better than Google’ (a company who’s whole business model centers on violating your privacy, particularly considering that you have the choice of paying if you strongly prefer not to have your data used to train the model.
I wouldn’t call DeepL “private” but I also wouldn’t consider them “privacy invasive”