I would be prepared to switch over completely for the languages they support if DeepL didn't insert their advertisement at the bottom of some results (I've had this accidentally included in an email before), and if they had a mobile app with a text selection action on Android to highlight text and see a popup with the translation. As it is, I use it most of the time but still rely on Google Translate being more accessible. Obviously DeepL is newer, so these limitations are understandable.

> if they had a mobile app with a text selection action on Android to highlight text and see a popup with the translation

I've had the same frustration, so I've written a simple script/app with Tasker (and AutoShare plugin) to select text and translate from DE -> EN. I use it quite frequently to learn German.

One weird quirk of DeepL is that when you become a $5/mo subscriber, you can't use their free /translator page while logged in.

Same here, I've been consistently surprised at how much better (and more natural/accurate) deepl is than google at translating, especially for sentences and paragraphs.

I love the feature that allows changing words with an alternative, which triggers a new translation with the new context.

I agree DeepL is often as good or better. (I will sometimes try both and compare. Usually I prefer about half from each, but both are fine on their own.) But DeepL doesn't support nearly as as many languages. Google Translate supports, to pick a few, Basque, Gujarati, Kazakh, Igbo and Thai. Some of those don't have a huge number of speakers, but they also have Arabic, Bengali, Hindi, Indonesian and Swahili, all in the top 20 by number of speakers.

Yes, I know some of these are done through a intermediary (often English, but not always). So if you translate Ukrainian → Catalan it probably really does Ukrainian → Russian → Spanish → Catalan. But it can still be helpful to have the options.

Is there any way to find out what the translation path is that GTranslate takes?

Background: I usually find I get better results when I translate from English to a random language X instead of from Y to X -- which is probably because it actually does Y -> English -> X. But I know more languages than Y and English, so it might be helpful to know what the "best" intermediary is, so to speak.

Not that I know of, except by trying to find mistakes like this:

https://languagelog.ldc.upenn.edu/nll/?p=47635

> Google translates the Japanese word "migaku 磨く" (to polish, to brush) to the French word "polonais" (Polish, as in "of Poland").

Meaning Japanese to French must go through English and get confused by Polish/polish.

Anecdote disclaimer.

Learning Spanish right now so I'm excited to learn about this better translator. The very first word I tested was "por favor".

DeepL detected it as English and translated it into Spanish to give "por favor" == "por favor".

Same for "una mas". To be fair, it worked once I properly accented the "a".

I'll be using this for longer sentences instead.