Thanks for getting back so quickly. I’ll gladly add you.
Think I’ve seen at least one other person complaining of this audio problem in another thread. Probably not as widespread as the text issue though, as you say.
Here’s hoping you have a eureka moment! 
Appreciate your time given to helping everyone here.
it is a quite widespread issue… i try to have natural voice all along my own and those i’m curating courses, but at the beginning i had only tts available, which in time were replaced… the tts still appears in listening skills, not for the right answer, but for the others…
i would not ask for solution, because, really… this is it, this whole thing, “it”, it is not worthy the effort of writing yet another script…
Okay, thanks 
After you’ve added me as a contributor, can you please send me a link to one of the levels where you’re seeing the problem?
PM-ed you btw 
Hi @neoncube,
over the past few weeks I have made extensive changes, i. e. prefixed verbs with “to” and nouns with “a”/“an”, made corrections and additions, deleted entries and so forth. I’ve finished my revisions today.
So I went ahead and threw your script several times at the course: Advanced English (C1, C2) - by csaba.olah.942 - Memrise
I’ve seen a few entries that turned gray. After three runs, the following entries were still gray instead of green:
OTOH, the following turned green during the third run after having been gray after the previous two runs:
I don’t know exactly what significance the gray entries have, but I suppose that these couldn’t be fixed?
Also, I have seen phantom entries in the course prior to the today’s changes. Is it that this attempt is just that and that it doesn’t help to execute the script consecutively?
LMK if I should you as a contributor to the course - in case you’d like to do some tests. Just note that there is almost 300k people that are or have been working on the course, so I’m a bit wary regarding experiments. 
Hi again @sarashaath. I’m afraid I may not have time to work on this any time soon, sorry 
I do programming for my day job, so it’s rare that I come home and do more programming just for fun (although it does happen sometimes! )
I also currently haven’t thought of a great way to solve the phantom audio issue. The only approaches that I’ve thought of so far come with risk, because they involve either uploading a fake audio entry and then immediately deleting it (which is risky because if the script dies between uploading the fake audio entry and deleting it, then there’s a fake audio entry left attached to the word) or deleting an existing audio entry and the quickly reuploading it (which sounds even more risky).
@duaal’s explanation of the gray entries is correct.
Are you getting red entries (signifying failures) sometimes when you run the script on a long course? If so, there might be ways to mitigate that.
I can’t think of a reason why entries would turn green after being gray during the previous runs, unless those items were modified, or some of the levels failed to expand when the script ran (which does happen sometimes, I think )
Totally understand. 
Thanks, @duaal and @neoncube! I suppose I didn’t search thoroughly enough as I actually did perform a search for gray here on the thread. 
Using the script in order to check for dupes is ideal too and adds another layer of usefulness! There was a whole bunch of them in the course, but I knew that I probably didn’t find them all. The fact that some entries vanished from the “gray list” may well be connected to changes that I performed between script runs.
I’ll check back on the course’s data next week and will let you know here.
Cheers!
@Olaf.Rabbachin No worries, and good to hear!
As a side note, dupes aren’t necessarily bad, depending on how the course is set up. Memrise is smart enough that if you learn an item in one level and then encounter it again in another level, it’ll already be marked as learned, and sometimes it can be useful to have the same item in multiple levels (if the item is a component in multiple dwords that are in different levels, for instance )
I think that only applies if the entries are 100% equal. In the aforementioned course the source translations differed though.
Just to update on the phantom audio issue - I realised the phantom audio entries that were bothering me were not quite as widespread as I thought. Only a few levels were affected so following @neoncube’s theory about deleting and reuploading, I did this manually and it worked a charm. 
I wonder if that’s why I have old definition bugs, like partially ( brackets. Or like, sometimes when I accidentally forget switch to commas rather than tab in bulk addition.
Hi @neoncube
I’ve tried the script today but the phantom entries are still there. I’ve deleted a whole level recently, ran the script several times (which all of the entries turned green) and tried the course. But most of the answers still came from that deleted level. Any idea on how I can fix them?
Hi @Cikinnad,
I believe Memrise draws possible answers from the course’s database, not directly from the other levels, so if the level was deleted but the entries in the database weren’t, then I think that could cause this issue 
Yea one example is when it was drawing answers from a sense deleted course section (I’m going to be bringing it back at the end of the course) and the fictional history of how the language developed.
Specifically how French and Japanese ran together in the US.
So you’ll have things like: the great fire that started the rebellion, presented as an answer to what Le Orenji is.
Because oranges and mutinys, absolutely related.
Thanks @neoncube
Still trying to sort things out. So many unwanted old entries in my course database