TTS Util does not collect any personal information. All information entered into or otherwise used by the app is kept on the device and is not transmitted.
Furthermore, the read/write storage permissions the app asks for are required only for being able to read selected files with text-to-speech or write synthesised speech into a new wave file.
Log data is collected in the case of errors in the app, such as crashes or situations where the app stops responding. This log data may include information such as your device’s Internet Protocol (“IP”) address, device name, operating system version, app configuration when the error occurred, the time and date of the error occurrence, the stack trace, and other statistics.
Log data collected from the app is useful in diagnosing why errors occur and will help to improve the experience of using the app overall. However, reporting errors is completely voluntary and you are free, of course, to not report errors when prompted after an app crash.
You can see which TTS engine you're using by navigating to your device's text-to-speech output settings, typically under Settings -> Accessibility -> Text-to-speech output.
Data processed by the selected text-to-speech engine may or may not be disclosed to outside parties. Once again, the TTS Util app has no control over this.
If the privacy of input text is very important to you, then Pico TTS, which is usually distributed with devices, only seems to require network access for downloading the initial models required for text synthesis. CMU Flite is another open source alternative that works like this.
This app is free software licensed under the Apache 2.0 license. What this means is that the source code is available. You can understand how it works, verify that my claims above are true and use parts of it in other apps if you want.
If there are any policy changes they will be posted on this page.