What do you dislike about ElevenLabs?
1) Voice consistency across models
When I train a voice in one model and then try to use it with a younger model, the voice changes quite noticeably and no longer sounds like the person I trained. Maintaining the identity of the original voice is very important for my projects, so this shift makes it difficult to use different models.
2) Pronunciation of numbers / “3D-style” words
The trained voice often struggles with numbers or terms like “3D.” The delivery feels unnatural and sometimes includes strange pauses in the middle of the phrase.
3) Emotional variation on first generation
When I generate a text for the first time, the read can sound flat. But if I regenerate the exact same text again, it suddenly becomes more fluent, expressive, and dynamic. I would really love to get that level of quality on the first try, without needing multiple regenerations.
4) Very important – incomplete dubbing / untranslated parts
During dubbing, some words or short segments are occasionally left in the original language or not dubbed at all. This is critical for me, because I then have to cut those parts out in my editing project and re-dub them separately, which slows everything down a lot.
I hope this feedback is helpful. I truly enjoy using ElevenLabs and would love to see improvements in these areas, as they would make a huge difference in my daily work.
Thank you for your time and support. Review collected by and hosted on G2.com.