We’re excited to announce that we’re launching a new and improved version of Automatic Text Capture (ATC) to localize your experiences. Over the course of the next few days, we will be slowly rolling out this new version of ATC.
While ATC is useful in speeding up the translation process, we know that it sometimes produces inconsistent results, so we wanted to improve it.
Some benefits include:
Improved Coverage
We are updating the way ATC scrapes experiences for strings. In addition to the real-time ingestion system that is used today, we are introducing an offline process which will check and update localization tables once a day. This will catch any strings missed or incorrectly scraped during the real-time process, which will greatly improve coverage and accuracy.
Localization Table Cleanup
ATC will also now check to see if there are strings that are no longer used in your experience and remove them to clean up your localization table. This could include strings that were mistakenly scraped, like player usernames, or temporary strings, like strings from special events. This will only apply to auto-scraped strings with automatic translations. Manually added strings and strings with manual translations will not be impacted.
For experiences that have ATC turned on already, the new version will be enabled by default. It can be turned off by going to your localization settings and disabling the second setting under “Automatic Text Capture”:
This is great! Can’t tell you how many times I’ve forgotten to turn off auto-localization on player labels or similar things. Obviously will still try to remember but it’s nice to have this as a backup.
I do have a couple questions about the offline scraping - does this happen on published content only or will work-in-progress content also be scraped and processed? Does it happen on all experiences or is there something like an engagement threshold that needs to be met?
Hi, this is a very good update, we were honestly really looking forward to it as there were a lot of translation issues, but the question is where is it?
We have localization tables right now in our game that are generated on run-time. This helps cover some very specific edge cases in our UI. However, we manually upload these localization tables which makes release really painful.
What’s the chance we can get a checkbox or something so that Roblox “scrapes” that localization table too. This would save us a serious amount of pain!
As someone in the thread already said, we’re still rolling this feature out. We’re taking it slow to make sure everything looks good - we’re aiming to have it rolled out to all experiences by the end of next week!
This issue should be largely be addressed, although there will be some edge cases that we still need to work on. The original process only sees content which is not translated (so in the linked example, it considered “Mori” already translated and only scraped the latter half of the word), but the new process looks at the entire source string.
Well, auto translate NEVER WORKED in my experience… and I’ve already tried resetting the languages and talking to Roblox support but the only thing they do is send links to the platform’s documentation but there is nothing to solve this problem, I’ve already tried to modify the description and title of the game and it doesn’t help, and it even seems like a supposed “shadow ban” because I tested the title of the experience and description in other experiences and it always translated normally, what do you think??
And I forgot to mention, auto translate is only working for strings but it doesn’t work for the title and description
Hey @GUILHERMENUMERO3 - sorry you’ve been having trouble with automatic translations. If you’d like to DM me the universe ID for your experience, we can investigate in the next 1-2 weeks and see if we can help you out.
