for TTS service using model aura-2-thalia-en Based on my observation, it seems that any word written in the format :MY_WORD: is being skipped randomly. what could be reason?
#TTS using model `aura-2-thalia-en` skipping word randomly.
humble mason
stoic trellisBOT
Hi @humble mason. I'm kapa.ai, your friendly support bot. I will try to answer your question by searching through the knowledge base. :wave:
Based on the provided knowledge sources, there is no explicit documentation or official guidance regarding the behavior of the Deepgram TTS model (including aura-2-thalia-en) when encountering words or phrases formatted as :<name>:. The available documentation focuses on best practices for punctuation, pauses, filler words, and pronunciation, but does not mention any reserved or special handling for colon-wrapped tokens or names.
However, the formatting guide for Aura-2 does emphasize that the quality and naturalness of the output are highly dependent on the input text formatting, and it warns against unexplained abbreviations, inconsistent formatting, and other non-standard text patterns, as these can negatively impact the model's interpretation and vocalization of the content. It is possible that using non-standard formats like :<name>: could cause the model to misinterpret or skip those tokens, especially if they are not typical in conversational or natural language input. The guide suggests reading your text aloud and formatting it as you would naturally speak, which may help avoid such issues Formatting Text for Aura-2.
In summary, while there is no direct mention of the :<name>: format causing words to be skipped, non-standard or unclear formatting can lead to unpredictable results in TTS output. Consider reformatting your input to use more natural language conventions and avoid special character patterns that are not documented as supported.