There's a difference between speed & latency when it comes to performance.
Faster-whisper can process audio faster than real-time, but AFAIK vanilla Whisper needs a few seconds long audio "frame" to do inference/STT.
Whisper Live fixes that, and reduces latency to a few tens/hundreds of ms.
