| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
| |
Bug: 5010576
Change-Id: Ia539a6a831a7d500d0c5336438e0bc0ff5854bf0
|
|\ |
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
This is one of 10 million errors (really) reported
by valgrind. I verified that some of them are OK, but it's
probably not worth going through the rest.
I expect most of those are just optimized code and
fancy bit trickery.
Change-Id: I3b8675ec8267691a5e6127a8e03239547107053e
|
|/
|
|
| |
Change-Id: I852d0619d0f5bcb7218d3ab266ae86d117bc95e7
|
|
|
|
| |
Change-Id: I6b5b81dc2c18d8a2b96074080304bfbdbc479fdc
|
|
|
|
| |
Change-Id: Ied389be15bfcd1a1c3ed289a0b1d6c7f7913629a
|
|
|
|
|
| |
bug:4476001
Change-Id: I14f4c5d803fc95bcaf11bc3ac182df4133083281
|
|
|
|
|
|
|
|
| |
This allows folks to depend on libttspico_static and just
call getTtsEngine() directly instead of all the dlopen()
dlsym() business.
Change-Id: I7b5b6c94d08b95ade9ba9c995e4bd2a862acdfb6
|
|
|
|
|
|
|
|
|
| |
The TTS engine should use only TTS defined constants, and
not constants from the audiosystem. This happened to work
because PCM_16_BIT in native was the same as DEFAULT in
java, which happened to be PCM_16_BIT.
Change-Id: I0f43a46afff8c45d1eb18c1beaab62dce87b8055
|
|
|
|
|
| |
Change-Id: I3bf2e6b3a0434f525dc1994f080f027482142ffd
Signed-off-by: Dima Zavin <dima@android.com>
|
|
|
|
|
| |
Change-Id: Id8e69601a0812a6a0034774adde672bb70662c6a
Signed-off-by: Dima Zavin <dima@android.com>
|
|
|
|
|
|
|
|
| |
Before, finalize() after shutdown() could segfault
since shutdown() kept the pointer to the deleted
SynthProxyJniStorage.
Change-Id: I338c19758733425be2e3b0d9e66176d11d27e8f6
|
|
|
|
| |
Change-Id: Ia634e6c84870ad6c2d1665da796adcd9cdc4ec8f
|
|
|
|
|
|
|
|
|
|
| |
The TTS engine now needs to have its own permissions
set, since it is running in its own process.
Also removes the minSdkVersion=2 to let the build system
set it automatically.
Change-Id: I7aea73ee5a415179f9b8b44e8f73c024b93f8983
|
|
|
|
| |
Change-Id: I4d94d41d3e5e9d9bd2b266ac4637c065ef13810a
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Requires TTS engine API added in
change I7614ff788e11f897e87052f684f1b4938d539fb7
The compatibility layer in
pico/compat/src/com/android/tts/compat/SynthProxy.java
pico/compat/jni/com_android_tts_compat_SynthProxy.cpp
is based on these files removed from the old TTS engine framework:
frameworks/base/packages/TtsService/src/android/tts/SynthProxy.java
frameworks/base/packages/TtsService/jni/android_tts_SynthProxy.cpp
Bug: 4150618
Change-Id: I7a2cca6b5cfbac6158a87fad69cc796140adb2f3
|
|
|
|
|
| |
Change-Id: I987cf0645cd1281e499426a29c82bdebd74f733b
Signed-off-by: Iliyan Malchev <malchev@google.com>
|
|\
| |
| |
| |
| | |
* commit 'baadee213d3350ab18fbd0c22a3ac3a65a853a4c':
Remove duplicate item to avoid compiling error
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The gcc 4.5 complains about the duplicate item:
external/svox/pico/lib/picosig2.c: In function 'sigAllocate':
external/svox/pico/lib/picosig2.c:75:26: error: operation on 'sig_inObj->int_vec38' may be undefined
Change-Id: I4532b770ff02e058253ed01cbed50b0efb8a83ce
|
| |
| |
| |
| | |
Change-Id: I3ce4ce7c121f3f72b242a0edf33b5f6f16e7fdac
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
If an utterance specifies its locale, the checkForLocale() function
looks for the index of the supported language that matches it. If
the exact locale is not supported, we look for language-only support.
But there could already be a match with the currently loaded language.
The fix consists in comparing the requested language with the one
that is loaded (if any), so we don't load a new one uselessly, AND
so the default language is used if it matches.
Change-Id: I501d946459ea982466eb804b17e9c43a8203564a
|
|\ \
| |/
| |
| |
| |
| | |
"Clean up makefiles for TTS data files."
Change-Id: I45a8f9887159de4d051455fc3feb202c26b9b8ce
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
This makes them look more like product definition files
that can better fit in a product inheritance chain.
Also, add a product definition file that includes all
languages (amusingly, that's the only combination
that's ever used).
Bug: 3065214
Change-Id: I42ef6038b7bda5a045562505e4bb2308e36b09f7
|
|/
|
|
|
|
|
|
|
|
|
|
| |
warning: operation on 'sig_inObj->int_vec38' may be undefined [-Wsequence-point]
The source code has cyclic initilization: a = c = b = c. gcc-4.5 reports
a warning that c may be undefined. The patch is simply removing the redundant
"= c".
Tested the patch with all gcc versions.
Change-Id: I4a6e9ee22b41058a34b0a657fb60e78c4e8f3ebb
|
|
|
|
| |
Change-Id: I109b34edc12547c644f546c4aaabbc8536579d62
|
|
|
|
| |
Change-Id: I26894c0029690bca05baeaee5198589242026130
|
|
|
|
| |
Change-Id: I541f1721baf4dadef653268ab1bc206ad521f036
|
|
|
|
|
|
| |
to be installed individually.
Change-Id: I29f98ff65a3c8c90747f635d0ad3887ab27a65d7
|
|\
| |
| |
| | |
the voice data is coming from Environment and is not hardcoded as /sdcard/"
|
| |
| |
| |
| |
| |
| |
| | |
for the voice data is coming from Environment and is not hardcoded
as /sdcard/
Change-Id: Ic7b69b6e2839cea5a273efb85fc030a8689b6307
|
|/
|
|
|
|
| |
figure out the alternative path for the lingware data.
Change-Id: If83e5e55ee7a420a41747d97beb36612a0b5fa1e
|
|
|
|
|
|
| |
engine config data and not support synthesis of IPA-only content.
Change-Id: Ibf308a17fb477e72f68467508190ddd928fde6ac
|
|
|
|
|
|
|
|
|
|
|
|
| |
Prevent random utterances in Pico TTS when synthesizing
an empty string.
In same patch:
Automatic volume raising tags in the Manager Layer are now removed
NOTE : It is still possible to selectively raise the volume for
portions of the input text, by embedding volume tags in the text
itself. These tags will not be removed by the Manager Layer.
Change-Id: I56fafcf6c1cff986a7394526027534308f227431
|
|
|
|
|
|
| |
as it sounds more muffled, particularly with eng-USA
Change-Id: I77c366afb1926f045ef931d0d66ce6a0bb377893
|
|\
| |
| |
| | |
closing tags not interpreted in some cases where enclosed text contains a smaller-then sign - Discontinuities in energy transformation - Lexicon Access: entry with empty phonetic transcription mistaken as :G2P in seldom cases"
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Issues fixed since 1.1.0.0:
- closing tags not interpreted in some cases where enclosed text
contains a smaller-then sign
- Discontinuities in energy transformation
- Lexicon Access: entry with empty phonetic transcription mistaken
as :G2P in seldom cases
|
|\ \
| | |
| | |
| | | |
unavailable voices."
|
| |/
| |
| |
| | |
available and unavailable voices.
|
| |
| |
| |
| | |
corrected strings for proper synthesis.
|
| |
| |
| |
| |
| | |
Patch from SVOX to fix the noise corruption issue in es-ES.
Added new test file to track regressions that are not related to XSAMPA.
|
|/
|
|
|
|
|
|
|
|
|
|
|
|
| |
Test updates to reflect fixes (in comments) and expand tests to allow
direct comparison between XSAMPA version and text version.
Fixes:
- de-DE: v Wahl "va:l fixed
- en-GB: A: (US) long "lA:N fixed
- en-US: l long "lA:N fixed
- fr-FR:
H juin "ZHE~ fixed
l long "lO~ fixed
l long "lO~ (in a sentence) fixed
w coin "kwE~ fixed
|
|
|
|
|
|
|
|
|
|
|
| |
branches merge paths.
Patch from SVOX: new lingware files for all languages.
en-US and en-GB are now XSAMPA compatible. Updated respective
test files to include cross-english tests.
en-GB passes en-US tests
en-US passes en-GB tests except for the following diphtongs:
"I_@", "U_@" and "e_@"
|
|
|
|
| |
of the supported languages.
|
|
|
|
| |
report its language strings for each language.
|
|
|
|
| |
Bug: #2361749.
|
|
|
|
|
|
|
|
|
|
| |
The build system does not honor user tags on apps,
and setting it is misleading.
This removes the confusion by making the makefiles
behave like they read.
Change-Id: Idb2afba98635dd41092cad0eb7432516b048bf67
|
|
|
|
|
|
|
| |
Updated test results after testing with the existing
lingware files.
The transcription changes come from corrections in the Pico
documentation from SVOX.
|
|
|
|
|
|
| |
the rhotacized open-mid central phoneme (code 0x025D). This corrects
the issue where the "ur" sound in "fur" or "nurse" causes the Pico
engine to synthesize garbage when using IPA input.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
pronounced by the TTS engine, and some are causing the whole utterance
to be unintelligible.
This is NOT a code change, this is a change in the language files
for the SVOX Pico TTS engine for en-US. Contact at SVOX was unable
to say whether the lingware correction was in the language resource
files or in the tools to generate the lingware files.
These new files correct identified "pollution" issues, that is cases
where a problem synthesizing one part of the utterance affects the
rest of the content that would otherwise be synthesized correctly.
The updated test file for en-US reflects the corrected issues. More
tests are underway to validate the new files, but initial tests are
conclusive and would fix issues with applications that use phonemic
input for the TTS engine.
DrNo (David Sparks) is aware of this update, and which applications
use phonemic input.
|
|\
| |
| |
| |
| | |
* changes:
Integrate patch from SVOX to complete IPA to XSAMPA mapping.
|