Do you mean 5+ seconds to finish the entire text? Because ChatGPT's generation was always instant and fast for me until they had constant server overload from the traffic. The time it took to generate entire paragraphs was faster than any TTS reading it in 2x speed.

The slow response nowadays is just an issue stemming from too many people using it at the same time and prioritising the paid version over the free one. ChatGPT was already good in its response time during the first few weeks. But I've yet to hear a TTS that can generate audio right off the bat without waiting for a few seconds.