Second commit is to be removed.
Signed-off-by: Bernhard Kölbl besentv@gmail.com
-- v10: windows.media.speech: Implement Vosk create and release functions in the unixlib. windows.media.speech/tests: Allow the SpeechRecognizer creation to fail in Wine. windows.media.speech/tests: Get rid of duplicated hresult. windows.media.speech: Add unixlib stub. windows.media.speech: Add Vosk checks to autoconf.
From: Bernhard Kölbl besentv@gmail.com
Signed-off-by: Bernhard Kölbl besentv@gmail.com --- configure.ac | 12 +++++++++++- 1 file changed, 11 insertions(+), 1 deletion(-)
diff --git a/configure.ac b/configure.ac index 9ff7c5e8914..6d43268c8c9 100644 --- a/configure.ac +++ b/configure.ac @@ -59,6 +59,7 @@ AC_ARG_WITH(udev, AS_HELP_STRING([--without-udev],[do not use udev (plug an AC_ARG_WITH(unwind, AS_HELP_STRING([--without-unwind],[do not use the libunwind library (exception handling)])) AC_ARG_WITH(usb, AS_HELP_STRING([--without-usb],[do not use the libusb library])) AC_ARG_WITH(v4l2, AS_HELP_STRING([--without-v4l2],[do not use v4l2 (video capture)])) +AC_ARG_WITH(vosk, AS_HELP_STRING([--without-vosk],[do not use Vosk])) AC_ARG_WITH(vulkan, AS_HELP_STRING([--without-vulkan],[do not use Vulkan])) AC_ARG_WITH(xcomposite,AS_HELP_STRING([--without-xcomposite],[do not use the Xcomposite extension]), [if test "x$withval" = "xno"; then ac_cv_header_X11_extensions_Xcomposite_h=no; fi]) @@ -483,7 +484,8 @@ AC_CHECK_HEADERS(\ syscall.h \ utime.h \ valgrind/memcheck.h \ - valgrind/valgrind.h + valgrind/valgrind.h \ + vosk_api.h ) WINE_HEADER_MAJOR() AC_HEADER_STAT() @@ -1787,6 +1789,14 @@ then WINE_WARNING([No sound system was found. Windows applications will be silent.]) fi
+dnl **** Check for Vosk **** +if test x$with_vosk != xno +then + WINE_CHECK_SONAME(vosk,vosk_recognizer_new,[AC_SUBST(VOSK_LIBS,"-lvosk")],,) +fi +WINE_NOTICE_WITH(vosk,[test x$with_vosk != xno -a x$ac_cv_lib_soname_vosk != xyes], + [libvosk ${notice_platform}development files not found (or too old), speech recognition won't be supported.]) + dnl *** Check for Vulkan *** if test "x$with_vulkan" != "xno" then
From: Bernhard Kölbl besentv@gmail.com
windows.media.speech: Add unixlib stub.
Signed-off-by: Bernhard Kölbl besentv@gmail.com --- dlls/windows.media.speech/Makefile.in | 2 ++ dlls/windows.media.speech/main.c | 19 +++++++++++++++ dlls/windows.media.speech/private.h | 4 ++++ dlls/windows.media.speech/unixlib.h | 33 +++++++++++++++++++++++++++ 4 files changed, 58 insertions(+) create mode 100644 dlls/windows.media.speech/unixlib.h
diff --git a/dlls/windows.media.speech/Makefile.in b/dlls/windows.media.speech/Makefile.in index 10903cb1d7b..c06a142780b 100644 --- a/dlls/windows.media.speech/Makefile.in +++ b/dlls/windows.media.speech/Makefile.in @@ -1,5 +1,7 @@ MODULE = windows.media.speech.dll +UNIXLIB = windows.media.speech.so IMPORTS = combase uuid +UNIX_LIBS = $(VOSK_LIBS)
C_SRCS = \ async.c \ diff --git a/dlls/windows.media.speech/main.c b/dlls/windows.media.speech/main.c index e772a791588..a3eb980438a 100644 --- a/dlls/windows.media.speech/main.c +++ b/dlls/windows.media.speech/main.c @@ -20,9 +20,28 @@ #include "initguid.h" #include "private.h"
+#include "unixlib.h" + #include "wine/debug.h"
WINE_DEFAULT_DEBUG_CHANNEL(speech); +WINE_DECLARE_DEBUG_CHANNEL(winediag); + +BOOL WINAPI DllMain(HINSTANCE instance, DWORD reason, void *reserved) +{ + if (reason == DLL_PROCESS_ATTACH) + { + DisableThreadLibraryCalls(instance); + if (__wine_init_unix_call()) + { + ERR_(winediag)("Wine is unable to load the Unix side dependencies for speech recognition. " + "Make sure Vosk is installed on your system and try again.\n"); + return FALSE; + } + } + + return TRUE; +}
HRESULT WINAPI DllGetClassObject(REFCLSID clsid, REFIID riid, void **out) { diff --git a/dlls/windows.media.speech/private.h b/dlls/windows.media.speech/private.h index e80d73ec1fb..2f804fbf1a7 100644 --- a/dlls/windows.media.speech/private.h +++ b/dlls/windows.media.speech/private.h @@ -22,6 +22,10 @@
#include <stdarg.h>
+#include "ntstatus.h" +#define WIN32_NO_STATUS +#include "winerror.h" +#include "winternl.h" #define COBJMACROS #include "corerror.h" #include "windef.h" diff --git a/dlls/windows.media.speech/unixlib.h b/dlls/windows.media.speech/unixlib.h new file mode 100644 index 00000000000..5516b51d235 --- /dev/null +++ b/dlls/windows.media.speech/unixlib.h @@ -0,0 +1,33 @@ +/* + * Unix library interface for Windows.Media.Speech + * + * Copyright 2023 Bernhard Kölbl for CodeWeavers + * + * This library is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * This library is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with this library; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA + */ + +#ifndef __WINE_WINDOWS_MEDIA_SPEECH_UNIXLIB_H +#define __WINE_WINDOWS_MEDIA_SPEECH_UNIXLIB_H + +#include <stdbool.h> +#include <stdint.h> + +#include "windef.h" +#include "winternl.h" +#include "wtypes.h" + +#include "wine/unixlib.h" + +#endif
From: Bernhard Kölbl besentv@gmail.com
Signed-off-by: Bernhard Kölbl besentv@gmail.com --- dlls/windows.media.speech/tests/speech.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-)
diff --git a/dlls/windows.media.speech/tests/speech.c b/dlls/windows.media.speech/tests/speech.c index 6bc5a8b1751..c7c7b6eb040 100644 --- a/dlls/windows.media.speech/tests/speech.c +++ b/dlls/windows.media.speech/tests/speech.c @@ -42,7 +42,6 @@ #define AsyncStatus_Closed 4
#define SPERR_WINRT_INTERNAL_ERROR 0x800455a0 -#define SPERR_WINRT_INCORRECT_FORMAT 0x80131537
#define IHandler_RecognitionResult ITypedEventHandler_SpeechContinuousRecognitionSession_SpeechContinuousRecognitionResultGeneratedEventArgs #define IHandler_RecognitionResultVtbl ITypedEventHandler_SpeechContinuousRecognitionSession_SpeechContinuousRecognitionResultGeneratedEventArgsVtbl @@ -1005,7 +1004,7 @@ static void test_SpeechSynthesizer(void) operation_ss_stream = (void *)0xdeadbeef; hr = ISpeechSynthesizer_SynthesizeSsmlToStreamAsync(synthesizer, str, &operation_ss_stream); /* Broken on Win 8 + 8.1 */ - ok(hr == S_OK || broken(hr == SPERR_WINRT_INCORRECT_FORMAT), "ISpeechSynthesizer_SynthesizeSsmlToStreamAsync failed, hr %#lx\n", hr); + ok(hr == S_OK || broken(hr == COR_E_FORMAT), "ISpeechSynthesizer_SynthesizeSsmlToStreamAsync failed, hr %#lx\n", hr);
if (hr == S_OK) {
From: Bernhard Kölbl besentv@gmail.com
To allow for error handling of missing Unix-side dependencies.
Signed-off-by: Bernhard Kölbl besentv@gmail.com --- dlls/windows.media.speech/tests/speech.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-)
diff --git a/dlls/windows.media.speech/tests/speech.c b/dlls/windows.media.speech/tests/speech.c index c7c7b6eb040..57e216b78d5 100644 --- a/dlls/windows.media.speech/tests/speech.c +++ b/dlls/windows.media.speech/tests/speech.c @@ -1192,7 +1192,7 @@ static void test_SpeechRecognizer(void) ok(ref == 1, "Got unexpected ref %lu.\n", ref);
hr = RoActivateInstance(hstr, &inspectable); - ok(hr == S_OK || broken(hr == SPERR_WINRT_INTERNAL_ERROR), "Got unexpected hr %#lx.\n", hr); + ok(hr == S_OK || hr == SPERR_WINRT_INTERNAL_ERROR, "Got unexpected hr %#lx.\n", hr);
if (hr == S_OK) { @@ -1411,7 +1411,7 @@ skip_operation: } else if (hr == SPERR_WINRT_INTERNAL_ERROR) /* Not sure when this triggers. Probably if a language pack is not installed. */ { - win_skip("Could not init SpeechRecognizer with default language!\n"); + skip("Could not init SpeechRecognizer with default language!\n"); }
done: @@ -1651,12 +1651,12 @@ static void test_Recognition(void) ok(hr == S_OK, "WindowsCreateString failed, hr %#lx.\n", hr);
hr = RoActivateInstance(hstr, &inspectable); - ok(hr == S_OK || broken(hr == SPERR_WINRT_INTERNAL_ERROR || hr == REGDB_E_CLASSNOTREG), "Got unexpected hr %#lx.\n", hr); + ok(hr == S_OK || hr == SPERR_WINRT_INTERNAL_ERROR || broken(hr == REGDB_E_CLASSNOTREG), "Got unexpected hr %#lx.\n", hr); WindowsDeleteString(hstr);
- if (FAILED(hr)) /* Win 8 and 8.1 and Win10 without enabled SR. */ + if (FAILED(hr)) /* Win 8 and 8.1 and Win10 without enabled SR. Wine with missing Unix side dependencies. */ { - win_skip("SpeechRecognizer cannot be activated!\n"); + skip("SpeechRecognizer cannot be activated!\n"); goto done; }
From: Bernhard Kölbl besentv@gmail.com
Signed-off-by: Bernhard Kölbl besentv@gmail.com --- dlls/windows.media.speech/Makefile.in | 5 +- dlls/windows.media.speech/private.h | 3 + dlls/windows.media.speech/recognizer.c | 42 +++++ dlls/windows.media.speech/unixlib.c | 222 +++++++++++++++++++++++++ dlls/windows.media.speech/unixlib.h | 20 +++ 5 files changed, 290 insertions(+), 2 deletions(-) create mode 100644 dlls/windows.media.speech/unixlib.c
diff --git a/dlls/windows.media.speech/Makefile.in b/dlls/windows.media.speech/Makefile.in index c06a142780b..6a6eb6162cf 100644 --- a/dlls/windows.media.speech/Makefile.in +++ b/dlls/windows.media.speech/Makefile.in @@ -1,6 +1,6 @@ MODULE = windows.media.speech.dll UNIXLIB = windows.media.speech.so -IMPORTS = combase uuid +IMPORTS = combase uuid user32 UNIX_LIBS = $(VOSK_LIBS)
C_SRCS = \ @@ -10,6 +10,7 @@ C_SRCS = \ main.c \ recognizer.c \ synthesizer.c \ - vector.c + unixlib.c \ + vector.c \
IDL_SRCS = classes.idl diff --git a/dlls/windows.media.speech/private.h b/dlls/windows.media.speech/private.h index 2f804fbf1a7..62952478bdf 100644 --- a/dlls/windows.media.speech/private.h +++ b/dlls/windows.media.speech/private.h @@ -31,6 +31,7 @@ #include "windef.h" #include "winbase.h" #include "winstring.h" +#include "winuser.h" #include "objbase.h"
#include "activation.h" @@ -47,6 +48,8 @@
#include "wine/list.h"
+#define SPERR_WINRT_INTERNAL_ERROR 0x800455a0 + /* * * Windows.Media.SpeechRecognition diff --git a/dlls/windows.media.speech/recognizer.c b/dlls/windows.media.speech/recognizer.c index c2f386206b8..1e72ede0eff 100644 --- a/dlls/windows.media.speech/recognizer.c +++ b/dlls/windows.media.speech/recognizer.c @@ -25,6 +25,9 @@
#include "wine/debug.h"
+#include "unixlib.h" +#include "wine/unixlib.h" + WINE_DEFAULT_DEBUG_CHANNEL(speech);
/* @@ -171,6 +174,8 @@ struct session IAudioCaptureClient *capture_client; WAVEFORMATEX capture_wfx;
+ speech_recognizer_handle unix_handle; + HANDLE worker_thread, worker_control_event, audio_buf_event; BOOLEAN worker_running, worker_paused; CRITICAL_SECTION cs; @@ -318,7 +323,9 @@ static ULONG WINAPI session_AddRef( ISpeechContinuousRecognitionSession *iface ) static ULONG WINAPI session_Release( ISpeechContinuousRecognitionSession *iface ) { struct session *impl = impl_from_ISpeechContinuousRecognitionSession(iface); + struct speech_release_recognizer_params release_params; ULONG ref = InterlockedDecrement(&impl->ref); + TRACE("iface %p, ref %lu.\n", iface, ref);
if (!ref) @@ -344,6 +351,9 @@ static ULONG WINAPI session_Release( ISpeechContinuousRecognitionSession *iface impl->cs.DebugInfo->Spare[0] = 0; DeleteCriticalSection(&impl->cs);
+ release_params.handle = impl->unix_handle; + WINE_UNIX_CALL(unix_vosk_release, &release_params); + IVector_ISpeechRecognitionConstraint_Release(impl->constraints); free(impl); } @@ -1079,6 +1089,35 @@ cleanup: return hr; }
+static HRESULT recognizer_factory_create_vosk_instance( struct session *session ) +{ + struct speech_create_recognizer_params create_params = { 0 }; + WCHAR locale[LOCALE_NAME_MAX_LENGTH]; + NTSTATUS status; + INT len; + + if (!(len = GetUserDefaultLocaleName(locale, LOCALE_NAME_MAX_LENGTH))) + return E_FAIL; + + if (CharLowerBuffW(locale, len) != len) + return E_FAIL; + + if (!WideCharToMultiByte(CP_ACP, 0, locale, len, (CHAR *)create_params.locale, ARRAY_SIZE(create_params.locale), NULL, NULL)) + return HRESULT_FROM_WIN32(GetLastError()); + + create_params.sample_rate = (FLOAT)session->capture_wfx.nSamplesPerSec; + + if ((status = WINE_UNIX_CALL(unix_vosk_create, &create_params))) + { + ERR("Unable to create Vosk instance for locale %s, status %#lx. Speech recognition won't work.\n", debugstr_a(create_params.locale), status); + return SPERR_WINRT_INTERNAL_ERROR; + } + + session->unix_handle = create_params.handle; + + return S_OK; +} + static HRESULT WINAPI recognizer_factory_Create( ISpeechRecognizerFactory *iface, ILanguage *language, ISpeechRecognizer **speechrecognizer ) { struct recognizer *impl; @@ -1125,6 +1164,9 @@ static HRESULT WINAPI recognizer_factory_Create( ISpeechRecognizerFactory *iface if (FAILED(hr = recognizer_factory_create_audio_capture(session))) goto error;
+ if (FAILED(hr = recognizer_factory_create_vosk_instance(session))) + goto error; + InitializeCriticalSection(&session->cs); session->cs.DebugInfo->Spare[0] = (DWORD_PTR)(__FILE__ ": recognition_session.cs");
diff --git a/dlls/windows.media.speech/unixlib.c b/dlls/windows.media.speech/unixlib.c new file mode 100644 index 00000000000..866aab2fb27 --- /dev/null +++ b/dlls/windows.media.speech/unixlib.c @@ -0,0 +1,222 @@ +/* + * Unixlib for Windows.Media.Speech + * + * Copyright 2023 Bernhard Kölbl for CodeWeavers + * + * This library is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * This library is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with this library; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA + */ + +#if 0 +#pragma makedep unix +#endif + +#include "config.h" + +#include <stdbool.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> +#include <wchar.h> + +#include <stdarg.h> +#include <dirent.h> +#include <dlfcn.h> +#include <errno.h> + +#ifdef HAVE_VOSK_API_H +#include <vosk_api.h> +#endif /* HAVE_VOSK_API_H */ + +#include "ntstatus.h" +#define WIN32_NO_STATUS +#include "winerror.h" +#include "winternl.h" + +#include "wine/debug.h" + +#include "unixlib.h" + +WINE_DEFAULT_DEBUG_CHANNEL(speech); + +#ifdef SONAME_LIBVOSK + +static inline speech_recognizer_handle vosk_recognizer_to_handle( VoskRecognizer *recognizer ) +{ + return (speech_recognizer_handle)(UINT_PTR)recognizer; +} + +static inline VoskRecognizer *vosk_recognizer_from_handle( speech_recognizer_handle handle ) +{ + return (VoskRecognizer *)(UINT_PTR)handle; +} + +static NTSTATUS find_model_by_locale_and_path( const char *path, const char *locale, VoskModel **model ) +{ + static const char *vosk_model_identifier_small = "vosk-model-small-"; + static const char *vosk_model_identifier = "vosk-model-"; + size_t ident_small_len = strlen(vosk_model_identifier_small); + size_t ident_len = strlen(vosk_model_identifier); + NTSTATUS status = STATUS_UNSUCCESSFUL; + char *dir_name, *model_path; + struct dirent *dirent; + size_t path_len; + DIR *dir; + + TRACE("path %s, locale %s, model %p.\n", path, debugstr_a(locale), model); + + if (!path || !locale || strlen(locale) < 4) + return STATUS_UNSUCCESSFUL; + + if (!(dir = opendir(path))) + return STATUS_UNSUCCESSFUL; + + path_len = strlen(path); + *model = NULL; + + while ((dirent = readdir(dir))) + { + if (dirent->d_type != DT_DIR) + continue; + + dir_name = dirent->d_name; + + if (!strncmp(dir_name, vosk_model_identifier_small, ident_small_len)) + dir_name += ident_small_len; + else if (!strncmp(dir_name, vosk_model_identifier, ident_len)) + dir_name += ident_len; + else + continue; + + /* First match for lang and region (en-us), then only lang (en). */ + if (strncmp(dir_name, locale, 5) && strncmp(dir_name, locale, 2)) + continue; + + if (!(model_path = malloc(path_len + 1 /* '/' */ + strlen(dirent->d_name) + 1))) + { + status = STATUS_NO_MEMORY; + break; + } + + sprintf(model_path, "%s/%s", path, dirent->d_name); + + TRACE("Trying to load Vosk model %s.\n", debugstr_a(model_path)); + + *model = vosk_model_new(model_path); + free(model_path); + + if (*model) + { + status = STATUS_SUCCESS; + break; + } + } + + closedir(dir); + + return status; +} + +static NTSTATUS find_model_by_locale( const char *locale, VoskModel **model ) +{ + const char *suffix = NULL; + char *env, *path = NULL; + NTSTATUS status; + + TRACE("locale %s, model %p.\n", debugstr_a(locale), model); + + if (!model) + return STATUS_UNSUCCESSFUL; + + if (!find_model_by_locale_and_path(getenv("VOSK_MODEL_PATH"), locale, model)) + return STATUS_SUCCESS; + if (!find_model_by_locale_and_path("/usr/share/vosk", locale, model)) + return STATUS_SUCCESS; + + if ((env = getenv("XDG_CACHE_HOME"))) + suffix = "/vosk"; + else if ((env = getenv("HOME"))) + suffix = "/.cache/vosk"; + else + return STATUS_UNSUCCESSFUL; + + if (!(path = malloc(strlen(env) + strlen(suffix) + 1))) + return STATUS_NO_MEMORY; + + sprintf(path, "%s%s", env, suffix); + status = find_model_by_locale_and_path(path, locale, model); + free(path); + + return status; +} + +static NTSTATUS speech_create_recognizer( void *args ) +{ + struct speech_create_recognizer_params *params = args; + VoskRecognizer *recognizer = NULL; + VoskModel *model = NULL; + NTSTATUS status = STATUS_SUCCESS; + + TRACE("args %p.\n", args); + + if ((status = find_model_by_locale(params->locale, &model))) + return status; + + if (!(recognizer = vosk_recognizer_new(model, params->sample_rate))) + status = STATUS_UNSUCCESSFUL; + + /* The model is kept alive inside the recognizer, so we can safely free our ref here. */ + vosk_model_free(model); + + params->handle = vosk_recognizer_to_handle(recognizer); + return status; +} + +static NTSTATUS speech_release_recognizer( void *args ) +{ + struct speech_release_recognizer_params *params = args; + + TRACE("args %p.\n", args); + + vosk_recognizer_free(vosk_recognizer_from_handle(params->handle)); + + return STATUS_SUCCESS; +} + +#else /* SONAME_LIBVOSK */ + +#define MAKE_UNSUPPORTED_FUNC( f ) \ + static NTSTATUS f( void *args ) \ + { \ + WARN("wine was compiled without Vosk support. Speech recognition won't work.\n"); \ + return STATUS_NOT_SUPPORTED; \ + } + +MAKE_UNSUPPORTED_FUNC(speech_create_recognizer) +MAKE_UNSUPPORTED_FUNC(speech_release_recognizer) +#undef MAKE_UNSUPPORTED_FUNC + +#endif /* SONAME_LIBVOSK */ + +unixlib_entry_t __wine_unix_call_funcs[] = +{ + speech_create_recognizer, + speech_release_recognizer, +}; + +unixlib_entry_t __wine_unix_call_wow64_funcs[] = +{ + speech_create_recognizer, + speech_release_recognizer, +}; diff --git a/dlls/windows.media.speech/unixlib.h b/dlls/windows.media.speech/unixlib.h index 5516b51d235..9dd5ac5ca28 100644 --- a/dlls/windows.media.speech/unixlib.h +++ b/dlls/windows.media.speech/unixlib.h @@ -30,4 +30,24 @@
#include "wine/unixlib.h"
+typedef UINT64 speech_recognizer_handle; + +struct speech_create_recognizer_params +{ + speech_recognizer_handle handle; + CHAR locale[LOCALE_NAME_MAX_LENGTH]; + FLOAT sample_rate; +}; + +struct speech_release_recognizer_params +{ + speech_recognizer_handle handle; +}; + +enum unix_funcs +{ + unix_vosk_create, + unix_vosk_release, +}; + #endif
Rémi Bernon (@rbernon) commented about dlls/windows.media.speech/recognizer.c:
if (CharLowerBuffW(locale, len) != len) return E_FAIL;
- if (!WideCharToMultiByte(CP_ACP, 0, locale, -1, (LPSTR)vosk_create_params.locale, len, NULL, NULL))
- if (!WideCharToMultiByte(CP_ACP, 0, locale, len, (CHAR *)create_params.locale, ARRAY_SIZE(create_params.locale), NULL, NULL))
I don't understand why you kept the cast here, I don't think you need it?
On Wed Feb 15 13:28:29 2023 +0000, Rémi Bernon wrote:
I don't understand why you kept the cast here, I don't think you need it?
Ah yeah, I had a brain fart. It was a leftover from an earlier warning.
On Wed Feb 15 08:55:04 2023 +0000, Bernhard Kölbl wrote:
It's also worth mentioning that I'm not going to use all of their functions. Generally I'd be more convinced if their library had a higher market share than it does now, but making their whole API more C compatible just for us and 3 used functions isn't really worth it to me.
I'm not advocating to improve their whole API; only modifying the parts we care about seems reasonable to me. The point is that if we have to make a change, I don't see why we should do it here instead of there.
On Wed Feb 15 18:49:32 2023 +0000, Zebediah Figura wrote:
I'm not advocating to improve their whole API; only modifying the parts we care about seems reasonable to me. The point is that if we have to make a change, I don't see why we should do it here instead of there.
I generally don't think you're wrong, but the tradeoff isn't visible to me given how little programs use Vosk, and even if they use it, it isn't via the C-API. Say the lib becomes more popular over time and speech recognition a less niche thing on Linux, I'm inclined to upstream it. Just keep in mind, what if we want to change to some different lib in the future? Additionally, I don't want to clutter their API with just some "C-Compatible"/non-Json functions.
-
Bernhard Kölbl -
Bernhard Kölbl (＠besentv)
-
Rémi Bernon (＠rbernon)
-
Zebediah Figura (＠zfigura)