MANY LIMITS and errors
Even if you have enough credits, it won't let you convert to audio if there are more than 10,000 characters of text. It also warns you that if you use more than 2,000 characters, the quality could be very poor. I finally cut the text to 9,990 characters, and the audio it gave me isn't correct. Although the pitch isn't bad, it's obvious from miles away that it's an AI that stops and pauses when it shouldn't and joins words together, even when the text is perfectly written specifically for an AI to convert it to audio. The same text has given me perfect results in Eleven Labs. I think it has potential, but it's not at the necessary level yet.
Bingchen_VoiSpark
Oct 1, 2025Thanks so much for taking the time to share your detailed feedback — it’s very helpful for us as we improve.
1/ Character limit: We completely understand your frustration. We’re already working on a long-input update that will allow users to enter more characters at once without splitting text.
2/ 2,000-character warning: This message only appears on Fish Audio and Orpheus models. In our internal tests, these two models often produced gibberish when handling more than 2,000 characters in a single input. To help prevent wasted credits, we added that warning. More advanced models don’t have this limitation.
3/ Pauses and flow: Sometimes models pause in unexpected places. You can improve this by using tools like break tags, or adding punctuation such as commas (,) or ellipses (...) to guide the voice.
Since VoiSpark also integrates ElevenLabs, you should be able to get the same results here as on the ElevenLabs platform when using their model. If you’re seeing a difference, we’d love to understand why. Please feel free to share details with us at contact@voispark.com.
Thanks again for trying VoiSpark and for helping us get better.