I am just seeing that new German voice Josef was added. This sounds so wrong and all the pauses at the wrong placec etc. Vicky in comparison sounds way more natural even though she is an older voice. Would a further refinement be coming up possibly?
When I created an automatic audio for a text, the text and the audio were not even synced meaning that the underlined move of the spontaneous words that were being spoken was not accurate…
I am also realizing that the audio was not created properly meaning the chunks of text was not simply read out loud and it was so bad that I just deleted the entire lesson and I never had this problem with older voices.
I just want to take a moment to emphasize once again that I am really expecting improved AI voices even for premium memberships that pay lower than the premium plus memberships thank you
Thanks, we’ll check this.
I have noticed a similar problem with the new voices in Japanese and Chinese. The pronunciation is very bad to the extent that I have to mute the audio.
The same happened with the new chinese voice: a lot of mistakes
Is there any update on when the new voice is actually going to be perfectly working because I feel like this is an important issue to tackle?
@kadirkaratas I asked our team to check Josef voice you mentioned and I’ll post here when I hear back from them.
Some days ago my app automatically switched to Josepf, but he’s an AI voice which is for Premium Plus only so it always defaulted to the really bad, robotic voice. I’ve since switched to Marlene and it’s much better.
I’m just noticing that there are three new voices with the AI symbol next to them, but I can’t actually use any of them.
I really hope that, as premium users, we’ll eventually get access to an AI voice and hopefully, it’s supposed to be just as good as, say, the existing Vicky voice and of course much better
@kadirkaratas AI voices are available in the Plus plan only.
Hi @kadirkaratas we made some adjustments. Could you please try to use Josef voice now? If it is still nok, could you please provide some lessons to check?
Thank you for providing the AI Josef voice. I just tried Josef for one sentence in sentence mode. Sounded good and natural. I then tried for a journal entry I wrote. The voice sounds natural…and I think sentence by sentence (in sentence mode) it would be ok, however, when doing a full lesson in normal mode, it sounds kind of funny. It’ll emphasize certain words that don’t really make sense to emphasize, or there will be a pause (seemingly also for emphasis), but it won’t really be in the right place. Almost like the “actor” is trying to overact each sentence on its own…out of context from the actual entire story. No idea what can or can’t be done to “tune” that.
It is definitely an improvement over TTS voice now pronunciation of words is more realistic. However, for a long lesson hearing this voice with those pauses in between will cause a headache. That’s the impression I have got .
First of all, I want to sincerely thank you for taking this issue seriously and for making an effort to improve the AI voice named Joseph. I truly appreciate that you’ve listened to the feedback and have been working on enhancing it.
That being said, I just tested the updated version of the voice with a seven-page lesson, and unfortunately, even within the first minute of the generated audio, I could still hear a lot of unnatural pauses. The rhythm and pacing of the speech just don’t flow naturally. It feels a bit mechanical, and the timing of the pauses often breaks the sense of continuity that you’d expect in normal spoken German.
I also want to emphasize that, in comparison, the automated female voice Viki still sounds far more natural and fluent. I can imagine that you’ve made some progress with Joseph — and perhaps there are improvements in certain sentence structures or tones — but overall, it’s still not at the same level as Viki. The biggest issue seems to be how Joseph handles intonation and rhythm. Sometimes, the voice speeds through a phrase too quickly, and then it suddenly pauses in odd places, creating a speech pattern that doesn’t align with how a real German speaker would naturally talk. The result feels quite disjointed, almost as if the AI doesn’t truly “understand” the language it’s speaking.
Again, I really want to thank you for your efforts and for being responsive to user feedback — that means a lot. But at this stage, I have to be honest and say that I’ll continue using the Viki voice for now, simply because it delivers a much more coherent and pleasant listening experience.
And while I don’t pretend to know much about the technical side of how these voices are developed, I can’t help but wonder if the solution might be a matter of fine-tuning the settings related to flow and pacing — perhaps something similar to how Microsoft Edge handles its AI voices in German. Their voices, for instance, sound incredibly fluent and human-like, with perfect balance between pauses, intonation, and natural speech rhythm.
Once again, I truly appreciate all the progress you’re making and the time you’re dedicating to refining Joseph’s voice. I’m confident that with some further adjustments, it could eventually become just as impressive as Viki — or maybe even better in the future.
@kadirkaratas thank you for your deep feedback. We are working on it 
BTW, when are you going to make AI voice in English for lifetime membership users available? I heard you are working on it. There are a few AI voices available but they come with a premium membership tier so I can not use it. It would improve my learning experience if you could make it avalable sooner rather than later. Just reminding you in any case.