Overview

Here you can view all currently known issues and possible solutions

The sophistication and advancement of artificial intelligence technology is amazing, but its operation often carries an elusive mystery. This is because its results often come from input data and are subject to the interpretation of model algorithms. To reduce this uncertainty, our team is committed to adding new features and continuously optimizing to make AI behavior more predictable and controllable. Nevertheless, when using generative AI, there are still some situations that may occur that need attention.

In this section, we will explore the challenges we have encountered in practice and issues reported by users. It should be understood that although potential problems are listed here, it does not mean that every user will encounter these troubles. In fact, in many cases, users can avoid these problems by taking appropriate preventive measures.

Poor Speech Generation Quality/Sudden Voice Change/Meaningless Content Generation

There may be multiple reasons for poor speech generation quality. Please refer to the possibilities in the following list:

• Character audio samples have quality issues, or AI model has insufficient learning of similar samples

• Generation parameter configuration has issues, or can be further improved

• Single sentence text is too long, or punctuation is too frequent

• Issues caused by current version model defects or limitations

Character Creation Failed

Currently, our models have different support capabilities for input audio samples. Before uploading, please check the sample language and ensure that your sample speech is greater than 2 seconds, and that the speech contains correctly identifiable speech content, and does not contain multiple speakers or complex multilingual content, otherwise it will cause character creation failure or other issues.

If you confirm that the speech itself has no problems, please try refreshing the page or trying again later.

Speech Generation Failed

V2 series version models only support Chinese and English. V3 version models add support for Cantonese, Japanese, Korean, French, German, Spanish and Portuguese, as well as more than 30 accent variants of these languages in total. Please ensure that the input text does not contain any other unsupported languages, and avoid mixing special symbols, content that does not belong to regular text (single punctuation marks, emoji symbols), otherwise it may cause generation failure and other issues.

If you confirm that the text itself has no problems, please try refreshing the page and trying again later. If you encounter failures multiple times, you can contact service support for feedback.