4 ms·
That's how cloud speech recognition is usually deployed. OpenAI whisper is faster than realtime on regular desktop CPUs, which I guess is good enough. And for
by fxtentacle 4y ago
That's how cloud speech recognition is usually deployed. OpenAI whisper is faster than realtime on regular desktop CPUs, which I guess is good enough.
And for a datacenter, a few $100 AMD CPUs will beat a single $20k NVIDIA A100 at throughput per dollar.