Rockbox mail archive
Subject: Re: Recorded voicefont file available (alpha)
From: Christi Scarborough (christi_at_chiark.greenend.org.uk)
> Yes, the file works. What a job, to record all these clipsy steady and
> straight, and trim them. Do you have any tool which assists this?
I used Audacity for the editing, recording the speech in a few passes
then joining it into one long data stream. I tried clipping
automatically with wavsilence, but it although it did a fairly good job
of the split, I'd have had to clip manually anyway, so I returned to
Audacity and selected and exported the WAV files from the initial stream.
> So far we've been using TTS, which is easy to reproduce and extend. But for
> sure a truly spoken file is a luxurious thing.
It's a fair amount of work, but being partially sighted, I was very
pleased to see Rockbox making the JB blind accessible, and I wanted to
add something to that effort. If someone has a soundcard with a low
noise mic line that they'd be willing to donate, I'd be willing to do it
again - it's a few hours work, but as you say, it's a nice luxury.
> The sound is hissing pretty bad (sorry to say), I personally woudn't
> use it.
I tried Audacity's built in noise removal facilities, but they made my
voice sound really weird and tinny. If anyone knows of a free tool with
better noise removal, please let me know. Of course the correct
solution is to get a better recording in the first place.
It seemed bearable to me listening on the PC, but people have different
standards, and it probably sounds different on the JB itself.
> I havent't seen a patch, but your IRC communication. IIRC voicefont.cpp so
> far just gathers all the files it can get. The selection which are non-empty
> and generate an entry was done before, in the TTS part. I agree that it
> makes sense to duplicate that detection.
The patch is at
It's a bit quick and dirty, but it works with the lang files as distributed.
> Thank you,
Page was last modified "Jan 10 2012" The Rockbox Crew