Discussion for huyvun's post

[regal gale]

@deep badge 🙌 Nice! Is this something you ever plan on publishing publicly for others to use?

[deep badge]

Yes, once I reduce the latency to ~300ms (currently >1500ms)..

[deep badge]

Carrying my Australian accent across in real-time,, hearing my own voice is really distracting though

[regal gale]

I think if you disable the microphone in bandicam (click the mic below record button), it shouldn't record your voice as you speak and only your Desktop audio. Aiseesoft Screen Recorder might be better, but if you have an Nvidia GPU then you should be using ShadowPlay. :3

[deep badge]

That's right, but I'm including my voice and the rendered voice so you can see the delay and compared how accurately the speech to speech is processed...

[regal gale]

Oh right- I thought you meant you didn't want to hear your voice in the recording not IRL. XD

Anyway- it sounds good and I'm looking forward to future developments!

[deep badge]

I didn't want to hear my voice because it's confusing me,, like hearing yourself echo on a long distance telephone call, causes you to stutter.. the solution was just to unplug my headphones while i was doing the s2s... once the delay is under 300ms , the confusion of hearing ones on voice will be pretty much eliminated.

[regal gale]

Shared it with my group who are also interested. Here's their first reactions to your first demo.

[deep badge]

Glad to see there is interest.. I wasn't sure if it would be useful to anyone. so yes, thank you Tybost. . I'll be posting update hopefully daily.

[tacit depot]

This is your project

[regal gale]

Oh yeah there's interest! I mean- I cant even conveniently use S2S on my browser of choice [Firefox] (recording mode has been broken for months, with no fix in sight) XD

[toxic wren]

FF issues no longer exist with this,.

[wide rampart]

Tagging this conversation so i can come back to it. Love seeing this work, thank you for sharing it with us!

In tests i was seeing about 400ms, which might be sufficient for say a streamer that is roleplaying to an audience? But right, have to turn both your own voice off and the rendered voice while doing this. It is hard enough hearing yourself.

I have not done this yet, but am considering ways to split the audio off to record for quality checks after the fact. This may be a low priority though as the streaming platforms do that for us already, and that is currently my audience.

[brisk field]

Hey i just joined ! I would love to try out that realtime performance ! @deep badge

[wide rampart]

That was the conversation actually, 400ms is not great for real-time conversations, But for one to many, that the translation is not yet real-time is not as large of a deal. Chat lags a bit, but other than that it seems to feel no worse than a broadcast situation (similar to terrestrial radio).

[brisk field]

you had the chance to test it ?

[brisk field]

seems like the person making it is not giving any more news, or not connecting to discord, quite scared that it won 't be finished..

[hallow berry]

lol please finish this project

[tacit depot]

Lol if he finish this let me know