I've already used a few TTS solutions (and yes, I miss SpeechKI). Overall, VoiSpark makes a solid impression, and the operation is simple and intuitive so far.

Voice cloning: Currently, this is certainly one of the most important features in TTS solutions. MiniMax's quality is good. I generally don't have a problem with the fact that the initial setup for a new cloned voice costs 100,000 credits. Cartesia and FishAudio didn't really convince me in my first test; those voices still sounded very unnatural.

Con:
What's a real challenge for me (and probably every other user) is that there's no preview option. There are many setting options, and you really have to test them all to achieve the best possible result. From the respective provider's model selection, to emotions, language boost, speed, volume, pitch, and even the "Text Normalization" function, you use up a lot of credits each time to eventually find the best setting.
Then comes the next challenge: Even though the AI ​​models (e.g., MiniMax) are now really good, the pronunciation of certain words can sometimes be off (I use it for German Voice overs). Currently, you can't preview or correct them. Here, too, you have to create the final voiceover each time (again, it costs a lot of credits) only to discover that it can't be used. This can quickly lead to frustration, especially with the lower tiers.

Overall, VoiSpark offers a solid foundation with good models. I sincerely hope that it will soon become significantly more user-friendly (preview and edit functions are a must).

Therefore, my rating fluctuates between 3 and 4 Tacos, but – with hope for future development – ​​I give it 4. However, using the voice cloning function can quickly become frustrating with Tiers 1 and 2.