Andrew, lets see if we can come up with a solution for tts in rb.
I will contact the author of espeak, and see if we can get him working with
us on this.
From: firstname.lastname@example.org [mailto:email@example.com] On
Behalf Of Andrew Hart
Sent: Friday, May 26, 2006 12:29 PM
Subject: RE: Tts in rockbox
This engine looks interesting. These comments are very
preliminary, as I've only spent five minutes looking at it.
Points in its favour:
Small; though I don't know what working space it needs to run.
It appears to be a phoneme-based tone-generated synthesizer which doesn't
require much memory or CPU grunt to run.
Points against it:
It currently dynamically links against the PortAudio library, which is most
probably unnecessary. It would be necessary to investigate if PortAudio is
really needed--my feeling is that it is not since Rockbox already has it's
own platform independent audio subsystem.
It's written in C++. It would have to be reworked in plain C so that it
could be compiled on all the different platforms independently and without
compatibility issues, as well as to conserve memory in the generated binary.
I need to study it some more. I have been looking at whether Flite would be
a possibility. Flite, however, is a concatinative voice synthesizer
requiring a reasonable amount of grunt. From what I can tell, Flite
requires 2-3 megs of static RAM or ROM to hold the software plus a meg of
working memory. The software codec capable players such as the Iriver H340
can probably handle Flite, but Flite is out of the question for the Archos
devices. Flite is also written in C.
At 15:36 26-05-2006, you wrote:
>It might take some work, but would this engine work?
>[mailto:firstname.lastname@example.org] On Behalf Of Jens Arnold
>Sent: Friday, May 26, 2006 11:06 AM
>Subject: Re: Tts in rockbox
>On 26.05.2006, Mark wrote:
> > None of the current targets are powerful enough to do understandable
> > TTS in realtime.
>This is imho not true. All our targets should have enough processing
>power to run realtime TTS. A plain Amiga 500 (7MHz MC68000, 512KB of
>RAM) was able to do it in the 80s, with a narrator.device of ~64KB and
>a translator.library of ~13KB.
>(Note: These are enhanced versions - the original AmigaOS 1.3 versions
>were probably still smaller)
>The biggest hurdle on archos was the inability to play PCM samples -
>but that's possible now thanks to the PCM codec.
>The real challenging task now would be to actually write a compact,
>efficient, opensource TTS engine, preferably supporting more than one
> > On Fri, 26 May 2006 13:39:15 +0100, Juan Hernandez <email@example.com>
> > wrote:
> >> Hello, I am blind, and would like to contribute to the rockbox
> >> project.
> >> I am wanting to write code for tts (Text to Speech) support in RB,
> >> right now, the voices idea works really well, but what if a blind
> >> user gets into an area that doesn't speak? I'd like to work on this.
> >> If any one has ideas on this please let me know.
Received on Sat May 27 00:26:41 2006