Trouble with speech recognition software, might be a sign that I lack something

I’ve been having trouble with my pronunciation.
Basically, I’ve been listening to Kenich repeating the same words so many times that I started immitating him when I practice speaking.
That quickly showed itself to be a problem when I started playing with voice recognition. It either picks up ん or んG where there should only be a G. The software works fine if I immitate Kyoko.

This must mean it’s not just the speech recognition software, but anyone would misunderstand me right?
Am I doing something wrong? How do I improve my pronunciation? (maybe Kenichi himself faces that same problem?)

Have you tried directly playing back Kenichi’s voice recordings for the voice recognition software and seeing if it can understand correctly? That should tell you if it’s you or Kenichi. Although I’m not sure how well playing a recording for voice recognition works. Just a thought, I don’t really know anything about voice recognition.


Speech recognition software in any language is notoriously bad at handling regional accents, soft (typically female) voices, and anything else other than talking like a television anchor. It’s not a reliable indicator of whether you would be understandable to a human.


To be fair, when I was starting, Google Assistant also had problems understanding me, but lately that’s less of a problem.

One thing to be cautious about is that the audio recordings on WaniKani are very exaggerated in terms of vowel length and there is a mix of standalone-ish and in-sentence clips from what I gathered, because words are pronounced with a slightly different melody when in context.

That being said, I tend to pronounce words more closely to Kyoko in general :sweat_smile:

That’s not necessarily on you, but if you would like to work on your pronunciation, try listening to news reports where announcers/anchors have to speak both fast and clearly to get the message across in allotted time.

1 Like

It’s just me. I even played it through my headphones and it worked fine, which are indeed of good quality but it’s still headphones.
Thanks, this really shows that I’m the problem.

Tokyo accent is the main accent of Japan I think

Which is what I do. It’s not like I can go any faster.

Is there a particular phrase you’re having trouble with? I tried my voice recognition with the nasalized-g and it picks it up fine.

1 Like

I must have been speaking it too slowly, because I tried mimmicking closer and closer and now it’s picking up just fine.
Although I did go back to the jisho one after I got the google one started picking it up normally and that one was just hopeless but I think it’s because the software is worse.

I’m still having other issues… It’s picking up 転がる as 頃ギャル (whatever that is) even if I don’t say the nasalized G, alongside other bizarre shenanigans. But I’m pretty sure this is not my fault.

Anyway, I regret making this thread. After you said you could do it just fine, I tried harder and harder and now I can get the Gs recognized just fine. I should have tried harder from the beginning. Sorry for the inconvenience.


It’s no trouble. :slight_smile:

As you’ve found, it’s really hard to tell when it’s not quite right without feedback. While voice recognition is one way to try yourself, it would be loads better to have an actual lesson with a Japanese speaker who can help you realtime.

1 Like

Tokyo dialect isn’t quite the same as standard Japanese, in much the same way that the London dialect isn’t quite the same as Recieved Pronunciaction (i.e. the “talking like a television anchor” mentioned above) is for British English.

This topic was automatically closed 365 days after the last reply. New replies are no longer allowed.