speechd-discuss
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: How to get rid of lengthy pause at end of sentences for speechSynthe


From: guest271314
Subject: Re: How to get rid of lengthy pause at end of sentences for speechSynthesis.speak()?
Date: Sat, 24 Aug 2024 14:45:46 -0700

>  then call the pipeline for the second sentence. Without this,
you'll have to wait for all the sentences to be synthesized before
getting any actual audio out.

Yes, that's the behaviour I am expecting.

> That's yet another reason for a C-based module for piper instead
of using generic, since in that case everything will be completely
pipelined.

Ideally piper, and all speech synthesis engines should be compiled to
WASM so we can run speech synthesis engines standalone and in the
browser using the same code, without any intermediary interface.

On Sat, Aug 24, 2024 at 2:37 PM Samuel Thibault
<samuel.thibault@ens-lyon.org> wrote:
>
> See how details do matter a lot...
>
> guest271314, le sam. 24 août 2024 14:23:54 -0700, a ecrit:
> > I posted the actual audio result. If you want I can use Bash instead
> > of Deno to execute pipe the same as Speech Dispatcher does.
> >
> > There is no next speech. I use two simple sentences with capital
> > letters starting each sentence and a period closing each sentence to
> > test.
>
> So you mean that you make just one call to the pipeline, providing the
> two sentences in one go?
>
> It happens that the generic module does split input at the delimiters,
> so as to pipeline the synthesis: call the pipeline for the first
> sentence, then call the pipeline for the second sentence. Without this,
> you'll have to wait for all the sentences to be synthesized before
> getting any actual audio out.
>
> If you really want to disable the pipelining, you can set
> GenericDelimiters to ""
>
> That's yet another reason for a C-based module for piper instead
> of using generic, since in that case everything will be completely
> pipelined.
>
> The generic approach is deemed to have such kinds of concerns since it's
> the inherent limitations of using an external pipeline that introduces
> such compromise between initial latency and pipelining.
>
> Samuel



reply via email to

[Prev in Thread] Current Thread [Next in Thread]