I want to manipulate live stream video before restreaming

I want to manipulate live stream video before restreaming. Basically i will use simple audio recognition then I manipulate video according to input then i restream.

CF Workers may work for it with Riconoscimento audio semplice: riconoscimento delle parole chiave  |  TensorFlow Core and if its integrated with stream i guess it may possible. Is that possible in Cloudflare? Might not? I accept a 3s delay due to workers if it’s not powerful. And what would be the 30h/month cost?