Real Time
Has anyone actually used this for Real Time (near realtime) translation yet? So far I have the M4t demo model up and running and I want to get cracking at the real time editor..
I wonder what's the best way to go about this.. open up loop with audio stream? output a stream of text? Process in chunks? Where would I separate audio chunks to pass the model?
I'm just thinking out loud.. Maybe i'll start and clip in 2 second chunk as I read this is the latency, but I feel it will lose the context from more complete chunks..
I'm interested in exploring this real-time (with slight delay) translation project as well. I have the M4T demo running and have tested it with different audio files that I have uploaded myself too. I'm thinking about a way to try the near real-time streaming project and was wondering how we can go about it.
Opening up a look with the audio stream? But doesn't the model take inputs in the form of saved .wav files? I'm finding it difficult to understand how we can start to figure this out