Really interesting presentation, AI agents possibilites give really nice possible improvements ! Good potential !
One or 2 questions to
@nikolascx if I may.
Regarding transcriptions, we can use 3CX Cloud up to 16SC 60k minutes /year.
We must then use local transcription thanks to a GPU-based component. Perfect.
Will this local / shared among various customers component need a specific licence (in addition to the ENT/AI licence on 3CX servers themselves), or will it run freely ?
Regarding its performances, you talk about 3 simultaneous calls at the same time (with a RTX 5090).
May be higher with DGX Spark you talk about
here.
But then, are calls transcribed at 1:1 speed ? Or is it faster ?
Finally, regarding AI agents.
You use OpenAI in the presentation, using GPT realtime for talking, and other models for reasoning.
Can't we use, or is it perhaps planned, to use the local GPU-based component to compute these tasks ?
Thank you for your precisions !