Replies: 1 comment
-
Why is this not the default procedure? |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
I tried Silero TTS and was annoyed that streaming output isn't shown live, so I took a look into extensions, and I seen that both SD, and TTS disable streaming during generation, as it would break them.
To fix it, I came up with an idea, to create a new hook, called for example
stream_modifier
, which would get called instead ofoutput_modifier
when streaming, and only when the generation has finished theoutput_modifier
would get called. I implemented a quick&dirty POC, and added it to Silero, however the full HTML reload of the entire history on every token looked bad (for example the audio players were flashing), so I decided to separate the current message from chat history. Right now, it looks like that:2023-04-18.03-42-51.mp4
You can try it out, by applying the following patch (on commit
ac2973f
): streaming.patch and then going to cai-chat (I broke the other chat modes), with just silero-tts enabled (I didn't test other extensions).If this idea gains support, and my way of implementing it isn't completly insane, I will bring it to PR stage when I have some time.
Beta Was this translation helpful? Give feedback.
All reactions