Subject: Feedback on Persian (Farsi) Language Accuracy - Request for Previous Model Access
Hi Tunee Team,
I am writing to share some feedback regarding the recent performance of the Persian (Farsi) language models on your platform.
As a frequent user, I noticed that the previous versions of Tunee were significantly more accurate and natural when processing Persian lyrics. In the current version, there is a noticeable decline in quality:
Pronunciation Issues: The AI now struggles with correct Persian phonetics and accents, which were handled perfectly in the past.
Vocal Naturalness: The voices have become more “robotic” and lose the emotional depth that the older model captured so well.
Rhythm Alignment: The timing between the Persian lyrics and the melody feels off compared to the older, more precise version.
Since the earlier model was much more reliable for Persian speakers, could you please provide an option to use the Legacy/Previous version? Or could your technical team look into the Persian linguistic processing of the current model to restore its former quality?
We love using Tunee and hope to see it return to its high standard of excellence for Farsi songs.
I want to report that I am having similar issues with Armenian language. The prior version was very accurate while everything generated in the past few days sounds like Japanese. At this point Tunee is not usable for me anymore. I need accurate Armenian vocals for my songs. Please go back to the previous version or find a fix. Thank you.
It’s not doing very good on English is either. That sounds crazy but there was an extreme drop in quality with the new update. The melody of the lyrics sounds extremely unnatural and is obviously AI. When it would sing lyrics before it sounded so natural and the melody flowed flawlessly. Now it mispronounces, has an unnatural melodic flow, and is just overall terrible.
It also can’t get the gender of the vocals correct. That’s crazy. You wonder why it doesn’t generate what you command but if something as clear as “female vocals” doesn’t get generated then it’s clearly disregarding most of what you tell it to do now. It’s a mess.
This topic was automatically closed after 24 hours. New replies are no longer allowed.