🌐 This article has been translated from Korean for your convenience. If there is any discrepancy, the Korean version prevails.
While using CharacterTalk, the character may occasionally fail to generate a normal response and instead repeatedly output the same or similar sentence as what you entered, or show other types of abnormal output. Below are the causes of this issue and what you can do about it.
What kind of behavior is this?
- When the character or story outputs exactly the same sentence as the one you entered
- When the same sentence or word is repeatedly output
- When responses are generated in an unintended foreign language (English, Japanese, Chinese, etc.) instead of the language you set
- When responses are generated that don’t fit the context or don’t make sense
- When the response gets cut off midway, or includes garbled characters or symbols
In addition to the examples above, various other forms of abnormal output may appear during the response-generation process.
Why does this happen?
CharacterTalk’s responses are generated using external generative AI technology. Because generative AI creates new text each time based on the input, it can occasionally produce various types of abnormal responses such as repeated or identical sentences to the input, unintended foreign languages, context errors, or interrupted output.
This is not an issue with a specific character or account, but a phenomenon that can commonly occur due to the technical characteristics of generative AI. POSTYPE is continuously working on improvements to enhance response quality.
Try this
If you encounter an abnormal response, the following methods may help.
- Try again: Even with the same input, trying to generate the response again often produces a normal answer.
- Rephrase your input: Rather than very short or vague input, including some situation and context in your sentence can help you receive more stable responses.
- Continue the conversation: Enter your next message so that it follows naturally from the previous flow of the conversation.
Can I get my points back?
Abnormal outputs that occur during the response-generation process—such as repeated sentences, unintended foreign languages, context errors, or interrupted output—are due to the characteristics of external generative AI technology. For this reason, we ask for your understanding that points used to generate such responses cannot be refunded.
These phenomena can occur not only in CharacterTalk but in any service that uses generative AI, and POSTYPE does not guarantee the accuracy or completeness of response results. However, we are continuously improving quality to provide a more stable experience. If the same issue occurs repeatedly or continuously, reporting it to our support team with the information below will greatly help our investigation.
- Time of occurrence
- Character used (character name or link)
- Frequency of occurrence
- Screenshots of the screen, if possible
Contact us
If it’s still not resolved after trying the methods above, or if you have additional questions, please contact our support team. We’ll check as quickly as possible and help you out.