I’ve seen this behavior intermittently for a while so I don’t think it’s a recent regression. I’m wondering how common it is, and what steps I could take to improve it. On my Voice Assistant Preview Edition Satellites, usually when the response is longer than average, the TTS starts, then pauses, then sometimes starts and pauses a few more times, and then usually ends before finishing. (If I were to make a blind guess based on the behavior I would wonder if something about the overlapping of the streaming response processing and TTS processing are maybe overloading the satellite CPU or at least running afoul of concurrency issues where the streaming response processing is starving the TTS processing?)
I don’t think I’m having a networking problem: I’m connected to a Ubiquiti network to a 2.4ghz only SSID broadcast from a single access point 6 feet away with a consistent signal strength better than -34dbm.
I’m using Nabu Casa STT and TTS, and this happens with any voice assistant (I’ve seen it with OpenAI direct, Gemini direct, and models on OpenRouter.)
Is this something others are seeing? Is it a known behavior? Is there troubleshooting I should do?
I’m absolutely loving voice assistant and I’d like to do a lot more with it. Getting the satellites to perform a bit better would be really awesome.