Re: GNU Guix 1.2.0rc1 available for testing!

[Ricardo Wurmus <rekado@elephly.net>]

Hello there,

Ludovic Courtès <ludo@gnu.org> writes:

> Tobias Platen <guix@platen-software.de> skribis:
>
>> Just saw that GNU Guix 1.2.0 released was recently released.
>> In the Ode to 1.2.0 the back vocals are done by Festival (likely singing-mode.scm).
>> Can I have the XML file that that was given to Festival.
>
> That’s a question for Ricardo.

I didn’t give it any XML file, but I can share my ~/.festivalrc:

--8<---------------cut here---------------start------------->8---
(Parameter.set 'Audio_Method 'Audio_Command)
;; This is "play" from the "sox" package.
(Parameter.set 'Audio_Command "play -q -t snd $FILE")
(Parameter.set 'Audio_Required_Format 'snd)
(Parameter.set 'Audio_Required_Rate 24000)

;; Alternatively, after "modprobe snd-pcm-oss" this would work:
;;(Parameter.set 'Audio_Method 'linux16audio)
;;(Parameter.set 'Audio_Device "/dev/dsp1")

(set! voice-path
      (cons "/home/rekado/vox/festival/lib/voices" voice-path))

(require 'voices)

;; Print all voices that have been found
(set! voice-location-trace t)
(search-for-voices)
(set! voice_default 'voice_cmu_us_fem_cg)
--8<---------------cut here---------------end--------------->8---

You may need to download the cmu_us_fem_cg voice data set from the
Festival website first.  I don’t think it’s included in the output of
the Festival installation in Guix.

With the lyrics in “poem.txt” you can then do: 

   text2wave -otype snd -o poem.wav poem.txt

I then chopped it up in Ardour to align individual chunks to the beat by
hand, applied some detuning with GxDetune (that’s why it sounds a little
like the Borg), and that’s pretty much it.  I added all these effects to
distract from the big quality difference that results from the low
sampling frequency of the generated file (I use 48kHz as the working
sampling frequency for all my audio projects).

I intend to publish more source files (e.g. Hydrogen patterns, Ardour
session file, etc) on my blog within the next month or so, perhaps even
including a Lilypond score (the output of the file I used had been
modified so much that I don’t think publishing it as is will be useful
to anyone).

-- 
Ricardo