Package: release.debian.org
Severity: normal
User: release.debian@packages.debian.org
Usertags: unblock
Hello,
espeak-ng used to have overlapping speech synthesis issues (#848016):
e.g. while moving fast in a list of items, the speech of each item would
get mixed with the previous one, making it difficult or impossible to
hear. We reduced the buffer size of espeak-ng from 200ms to 50ms to
considerably reduce the issue.
However, the modified buffer size happened to break the use of mbrola
voices (#860891), so we raised it a bit, from 50ms to 60ms, like
upstream did, to fix that.
However, that brought back some of the overlapping issues, making it
tedious to use...
(https://lists.debian.org/debian-accessibility/2017/06/msg2.html)
I have thus uploaded a version -11 of espeak-ng (attached debdiff) which
reduces it to 49ms, which avoids the overlapping regression, and which I
have tested as working with all mbrola voices.
unblock espeak-ng/1.49.0+dfsg-11
This contains udebs, so Cc-ing KiBi for the udeb ack.
-- System Information:
Debian Release: 9.0
APT prefers testing
APT policy: (990, 'testing'), (500, 'unstable-debug'), (500,
'testing-debug'), (500, 'buildd-unstable'), (500, 'unstable'), (500, 'stable'),
(500, 'oldstable'), (1, 'experimental-debug'), (1, 'buildd-experimental'), (1,
'experimental')
Architecture: amd64
(x86_64)
Foreign Architectures: i386
Kernel: Linux 4.11.0 (SMP w/4 CPU cores)
Locale: LANG=fr_FR.UTF-8, LC_CTYPE=fr_FR.UTF-8 (charmap=UTF-8)
Shell: /bin/sh linked to /bin/dash
Init: systemd (via /run/systemd/system)
--
Samuel
ya(ka|ma|to)* ca existe une fois sur 2 au japon, c'est facile ;-)
-+- #ens-mim au japon -+-
diff -Nru espeak-ng-1.49.0+dfsg/debian/changelog
espeak-ng-1.49.0+dfsg/debian/changelog
--- espeak-ng-1.49.0+dfsg/debian/changelog 2017-04-29 16:32:54.0
+0200
+++ espeak-ng-1.49.0+dfsg/debian/changelog 2017-06-05 22:04:57.0
+0200
@@ -1,3 +1,11 @@
+espeak-ng (1.49.0+dfsg-11) unstable; urgency=medium
+
+ * patches/bufsize: Increasing the buffersize to 60ms brought back some
+overlapping. Revert this to 49ms, which both avoids overlapping, and
+was tested to work fine with all MBROLA voices.
+
+ -- Samuel Thibault Mon, 05 Jun 2017 22:04:57 +0200
+
espeak-ng (1.49.0+dfsg-10) unstable; urgency=medium
* patches/bufsize: Increase buffersize to 60ms like upstream did, to fix
using
diff -Nru espeak-ng-1.49.0+dfsg/debian/patches/bufsize
espeak-ng-1.49.0+dfsg/debian/patches/bufsize
--- espeak-ng-1.49.0+dfsg/debian/patches/bufsize2017-04-29
16:32:54.0 +0200
+++ espeak-ng-1.49.0+dfsg/debian/patches/bufsize2017-06-05
22:04:57.0 +0200
@@ -5,7 +5,7 @@
// buflength is in mS, allocate 2 bytes per sample
if ((buffer_length == 0) || (output_mode & ENOUTPUT_MODE_SPEAK_AUDIO))
- buffer_length = 200;
-+ buffer_length = 60;
++ buffer_length = 49;
outbuf_size = (buffer_length * samplerate)/500;
out_start = (unsigned char *)realloc(outbuf, outbuf_size);