Actually I use the OpenAI API and TTS costs much more than text generation itself and of course it is not in the robots voice.
Support for XTTSv2 (xtts-api-server) would make it possible to use TTS locally and even train it with the Robots original voice for using it in many different languages should be possible as well. XTTSv2 itself supports many different languages while other TTS AIs only support english, can be trained by some less minutes audio material (if not a 6s sample is enough), is very quick in generation and the output is also solid.
Actually I use the OpenAI API and TTS costs much more than text generation itself and of course it is not in the robots voice.
Support for XTTSv2 (xtts-api-server) would make it possible to use TTS locally and even train it with the Robots original voice for using it in many different languages should be possible as well. XTTSv2 itself supports many different languages while other TTS AIs only support english, can be trained by some less minutes audio material (if not a 6s sample is enough), is very quick in generation and the output is also solid.