what is the best combination for low-latency?

[sergenti]

best TTS + STT pair?
is there one that has medium-high quality and a short time to wait between responses?

[lugia19]

In terms of TTS, the only real option is elevenlabs, but it's about as latency-optimized as it can be.
In terms of STT, I'd say local whisper (if your GPU is up to the task of running it).

The smaller the model, the faster the voice recognition runs, but the less accurate it is. For english, I've found the medium size works well enough.