Bump whisper, clblast, add buffered output

.github/scripts/Package-Windows.ps1

@@ -4,7 +4,7 @@ param(
     [string] $Target = 'x64',
     [ValidateSet('Debug', 'RelWithDebInfo', 'Release', 'MinSizeRel')]
     [string] $Configuration = 'RelWithDebInfo',
-    [ValidateSet('cpu', '12.2.0', '11.8.0')]
+    [ValidateSet('cpu', 'clblast', '12.2.0', '11.8.0')]
     [string] $Cublas = 'cpu',
     [switch] $BuildInstaller,
     [switch] $SkipDeps
@@ -52,6 +52,8 @@ function Package {
     # Check if $cublas is cpu or cuda
     if ( $Cublas -eq 'cpu' ) {
         $CudaName = 'cpu'
+    } elseif ( $Cublas -eq 'cblast' ) {
+        $CudaName = 'cblast'
     } else {
         $CudaName = "cuda${Cublas}"
     }

.github/workflows/build-project.yaml

@@ -246,7 +246,7 @@ jobs:
     needs: check-event
     strategy:
       matrix:
-        cublas: [cpu, 12.2.0, 11.8.0]
+        cublas: [cpu, clblast, 12.2.0, 11.8.0]
     defaults:
       run:
         shell: pwsh

.github/workflows/push.yaml

@@ -76,6 +76,7 @@ jobs:
 
           variants=(
             'windows-x64-cpu;zip|exe'
+            'windows-x64-clblast;zip|exe'
             'windows-x64-11.8.0;zip|exe'
             'windows-x64-12.2.0;zip|exe'
             'macos-arm64;tar.xz|pkg'

CMakeLists.txt

@@ -92,6 +92,7 @@ target_sources(
           src/whisper-utils/whisper-processing.cpp
           src/whisper-utils/whisper-utils.cpp
           src/whisper-utils/silero-vad-onnx.cpp
-          src/translation/translation.cpp)
+          src/translation/translation.cpp
+          src/utils.cpp)
 
 set_target_properties_plugin(${CMAKE_PROJECT_NAME} PROPERTIES OUTPUT_NAME ${_name})

cmake/BuildCTranslate2.cmake

@@ -21,10 +21,11 @@ elseif(WIN32)
 
   # check CPU_OR_CUDA environment variable
   if(NOT DEFINED ENV{CPU_OR_CUDA})
-    message(FATAL_ERROR "Please set the CPU_OR_CUDA environment variable to either CPU or CUDA")
+    message(
+      FATAL_ERROR "Please set the CPU_OR_CUDA environment variable to either `cpu`, `clblast`, `12.2.0` or `11.8.0`")
   endif()
 
-  if($ENV{CPU_OR_CUDA} STREQUAL "cpu")
+  if($ENV{CPU_OR_CUDA} STREQUAL "cpu" OR $ENV{CPU_OR_CUDA} STREQUAL "clblast")
     FetchContent_Declare(
       ctranslate2_fetch
       URL https://github.com/occ-ai/obs-ai-ctranslate2-dep/releases/download/1.2.0/libctranslate2-windows-4.1.1-Release-cpu.zip

cmake/BuildWhispercpp.cmake

@@ -14,12 +14,12 @@ if(APPLE)
   endif(NOT DEFINED ENV{MACOS_ARCH})
 
   set(WHISPER_CPP_URL
-      "https://github.com/occ-ai/occ-ai-dep-whispercpp/releases/download/0.0.1/whispercpp-macos-$ENV{MACOS_ARCH}-0.0.1.tar.gz"
+      "https://github.com/occ-ai/occ-ai-dep-whispercpp/releases/download/0.0.2/whispercpp-macos-$ENV{MACOS_ARCH}-0.0.2.tar.gz"
   )
   if($ENV{MACOS_ARCH} STREQUAL "x86_64")
-    set(WHISPER_CPP_HASH "36F39F02F999AAF157EAD3460DD00C8BDAA3D6C4A769A9E4F64E327871B4B11F")
+    set(WHISPER_CPP_HASH "00C308AF0BFFF7619934403A8080CC9AFC4EDAA328D7587E617150A2C6A33313")
   elseif($ENV{MACOS_ARCH} STREQUAL "arm64")
-    set(WHISPER_CPP_HASH "6AF7BB904B03B6208B4281D44465B727FB608A32CABD1394B727937C5F4828A1")
+    set(WHISPER_CPP_HASH "0478E2079E07FA81BEE77506101003F4A4C8F0DF9E23757BD7E1D25DCBD1DB30")
   else()
     message(
       FATAL_ERROR
@@ -45,24 +45,31 @@ elseif(WIN32)
   if(NOT DEFINED ENV{CPU_OR_CUDA})
     message(
       FATAL_ERROR
-        "The CPU_OR_CUDA environment variable is not set. Please set it to either `cpu` or `11.8.0` or `12.2.0`")
+        "The CPU_OR_CUDA environment variable is not set. Please set it to either `cpu`, `clblast` or `11.8.0` or `12.2.0`"
+    )
   endif(NOT DEFINED ENV{CPU_OR_CUDA})
 
-  set(CUDA_PREFIX $ENV{CPU_OR_CUDA})
-  if(NOT $ENV{CPU_OR_CUDA} STREQUAL "cpu")
-    set(CUDA_PREFIX "cuda$ENV{CPU_OR_CUDA}")
+  set(ARCH_PREFIX $ENV{CPU_OR_CUDA})
+  if(NOT $ENV{CPU_OR_CUDA} STREQUAL "cpu" AND NOT $ENV{CPU_OR_CUDA} STREQUAL "clblast")
+    set(ARCH_PREFIX "cuda$ENV{CPU_OR_CUDA}")
     add_compile_definitions("LOCALVOCAL_WITH_CUDA")
+  elseif($ENV{CPU_OR_CUDA} STREQUAL "cpu")
+    add_compile_definitions("LOCALVOCAL_WITH_CPU")
+  else()
+    add_compile_definitions("LOCALVOCAL_WITH_CLBLAST")
   endif()
 
   set(WHISPER_CPP_URL
-      "https://github.com/occ-ai/occ-ai-dep-whispercpp/releases/download/0.0.1/whispercpp-windows-${CUDA_PREFIX}-0.0.1.zip"
+      "https://github.com/occ-ai/occ-ai-dep-whispercpp/releases/download/0.0.2/whispercpp-windows-${ARCH_PREFIX}-0.0.2.zip"
   )
   if($ENV{CPU_OR_CUDA} STREQUAL "cpu")
-    set(WHISPER_CPP_HASH "5261FCCD18BA52AE7ECD37617452F0514238FAB4B12713F1FCA491F4ABA170AA")
+    set(WHISPER_CPP_HASH "6DE628A51B9352624A1EC397231591FA3370E6BB42D9364F4F91F11DD18F77D2")
+  elseif($ENV{CPU_OR_CUDA} STREQUAL "clblast")
+    set(WHISPER_CPP_HASH "97BF58520F1818B7C9F4E996197F3097934E5E0BBA92B0B016C6B28BE9FF1642")
   elseif($ENV{CPU_OR_CUDA} STREQUAL "12.2.0")
-    set(WHISPER_CPP_HASH "1966A6C7347FCB9529140F8097AED306F31C6DDE328836FD6498A980E20B8E6C")
+    set(WHISPER_CPP_HASH "48C059A3364E0AAD9FB0D4194BA554865928D22A27ECE5E3C116DC672D5D6EDE")
   elseif($ENV{CPU_OR_CUDA} STREQUAL "11.8.0")
-    set(WHISPER_CPP_HASH "172F4021E888A89A694373AE0888C04DB99BC11F3A2633270248E03AF5AC762E")
+    set(WHISPER_CPP_HASH "29A5530E83896DE207F0199535CBBB24DF0D63B1373BA66139AD240BA67120EB")
   else()
     message(
       FATAL_ERROR

data/locale/ar-SA.ini

@@ -49,3 +49,5 @@ source_language="لغة المصدر"
 translate="ترجمة (⚠️ زيادة المعالجة)"
 translate_add_context="الترجمة مع السياق"
 whisper_translate="ترجمة إلى الإنجليزية (Whisper)"
+buffer_size_msec="حجم الذاكرة المؤقتة (ملي ثانية)"
+overlap_size_msec="حجم التداخل (ملي ثانية)"

data/locale/de-DE.ini

@@ -49,3 +49,5 @@ source_language="Quellsprache"
 translate="Übersetzen (⚠️ erhöhte Verarbeitung)"
 translate_add_context="Mit Kontext übersetzen"
 whisper_translate="Ins Englische übersetzen (Flüstern)"
+buffer_size_msec="Puffergröße (ms)"
+overlap_size_msec="Überlappungsgröße (ms)"

data/locale/en-US.ini

@@ -12,7 +12,7 @@ text_file_output="Text File output"
 output_filename="Output filename"
 whisper_model="Whisper Model"
 external_model_file="External model file"
-whisper_parameters="Advanced Settings"
+whisper_parameters="Whisper Model Parameters"
 language="Language"
 whisper_sampling_method="Whisper Sampling Method"
 n_threads="Number of threads"
@@ -49,3 +49,5 @@ source_language="Source language"
 translate="Translate (⚠️ increased processing)"
 translate_add_context="Translate with context"
 whisper_translate="Translate to English (Whisper)"
+buffer_size_msec="Buffer size (ms)"
+overlap_size_msec="Overlap size (ms)"

data/locale/es-ES.ini

@@ -49,3 +49,5 @@ source_language="Idioma fuente"
 translate="Traducir (⚠️ procesamiento aumentado)"
 translate_add_context="Traducir con contexto"
 whisper_translate="Traducir al inglés (Whisper)"
+buffer_size_msec="Tamaño del búfer (ms)"
+overlap_size_msec="Tamaño de superposición (ms)"

data/locale/fr-FR.ini

@@ -49,3 +49,5 @@ source_language="Langue source"
 translate="Traduire (⚠️ traitement accru)"
 translate_add_context="Traduire avec contexte"
 whisper_translate="Traduire en anglais (Whisper)"
+buffer_size_msec="Taille du tampon (ms)"
+overlap_size_msec="Taille de chevauchement (ms)"

data/locale/hi-IN.ini

@@ -49,3 +49,5 @@ source_language="स्रोत भाषा"
 translate="अनुवाद करें (⚠️ बढ़ी प्रसंस्करण)"
 translate_add_context="संदर्भ के साथ अनुवाद करें"
 whisper_translate="अंग्रेजी में अनुवाद करें (व्हिस्पर)"
+buffer_size_msec="बफ़र आकार (ms)"
+overlap_size_msec="ओवरलैप आकार (ms)"

data/locale/ja-JP.ini

@@ -49,3 +49,5 @@ source_language="ソース言語"
 translate="翻訳 (⚠️処理増加)"
 translate_add_context="コンテキスト付きで翻訳"
 whisper_translate="英語に翻訳（ウィスパー）"
+buffer_size_msec="バッファサイズ（ms）"
+overlap_size_msec="オーバーラップサイズ（ms）"

data/locale/ko-KR.ini

@@ -49,3 +49,5 @@ source_language="원본 언어"
 translate="번역 (⚠️ 처리 시간 증가)"
 translate_add_context="컨텍스트와 함께 번역"
 whisper_translate="영어로 번역 (속삭임)"
+buffer_size_msec="버퍼 크기 (ms)"
+overlap_size_msec="오버랩 크기 (ms)"

data/locale/pl-PL.ini

@@ -49,3 +49,5 @@ source_language="Język źródłowy"
 translate="Tłumacz (⚠️ zwiększone przetwarzanie)"
 translate_add_context="Tłumacz z kontekstem"
 whisper_translate="Tłumacz na angielski (Whisper)"
+buffer_size_msec="Rozmiar bufora (ms)"
+overlap_size_msec="Rozmiar nakładki (ms)"

data/locale/pt-BR.ini

@@ -49,3 +49,5 @@ source_language="Língua de origem"
 translate="Traduzir (⚠️ o processamento aumentará)"
 translate_add_context="Traduzir com contexto"
 whisper_translate="Traduzir para inglês (Whisper)"
+buffer_size_msec="Tamanho do buffer (ms)"
+overlap_size_msec="Tamanho da sobreposição (ms)"

data/locale/ru-RU.ini

@@ -48,3 +48,5 @@ source_language="Исходный язык"
 translate="Перевести (⚠️ обработка будет увеличена)"
 translate_add_context="Перевести с контекстом"
 whisper_translate="Перевести на английский (Whisper)"
+buffer_size_msec="Размер буфера (мс)"
+overlap_size_msec="Размер перекрытия (мс)"

data/locale/zh-CN.ini

@@ -49,3 +49,5 @@ source_language="源语言"
 translate="翻译 （⚠️ 增加处理）"
 translate_add_context="带上下文翻译"
 whisper_translate="翻译为英语（Whisper）"
+buffer_size_msec="缓冲区大小（毫秒）"
+overlap_size_msec="重叠大小（毫秒）"

src/captions-thread.h

@@ -0,0 +1,118 @@
+#ifndef CAPTIONS_THREAD_H
+#define CAPTIONS_THREAD_H
+
+#include <queue>
+#include <vector>
+#include <chrono>
+#include <thread>
+#include <mutex>
+#include <condition_variable>
+#include <functional>
+#include <string>
+
+#include <obs.h>
+
+#include "plugin-support.h"
+
+class CaptionMonitor {
+public:
+	// default constructor
+	CaptionMonitor() = default;
+
+	~CaptionMonitor()
+	{
+		{
+			std::lock_guard<std::mutex> lock(queueMutex);
+			stop = true;
+		}
+		condVar.notify_all();
+		workerThread.join();
+	}
+
+	void initialize(std::function<void(const std::string &)> callback_, size_t maxSize_,
+			std::chrono::seconds maxTime_)
+	{
+		this->callback = callback_;
+		this->maxSize = maxSize_;
+		this->maxTime = maxTime_;
+		this->initialized = true;
+		this->workerThread = std::thread(&CaptionMonitor::monitor, this);
+	}
+
+	void addWords(const std::vector<std::string> &words)
+	{
+		{
+			std::lock_guard<std::mutex> lock(queueMutex);
+			for (const auto &word : words) {
+				wordQueue.push_back(word);
+			}
+			this->newDataAvailable = true;
+		}
+		condVar.notify_all();
+	}
+
+private:
+	void monitor()
+	{
+		obs_log(LOG_INFO, "CaptionMonitor::monitor");
+		auto startTime = std::chrono::steady_clock::now();
+		while (true) {
+			std::unique_lock<std::mutex> lock(this->queueMutex);
+			// wait for new data or stop signal
+			this->condVar.wait(lock,
+					   [this] { return this->newDataAvailable || this->stop; });
+
+			if (this->stop) {
+				break;
+			}
+
+			if (this->wordQueue.empty()) {
+				continue;
+			}
+
+			// emit up to maxSize words from the wordQueue
+			std::vector<std::string> emitted;
+			while (!this->wordQueue.empty() && emitted.size() <= this->maxSize) {
+				emitted.push_back(this->wordQueue.front());
+				this->wordQueue.pop_front();
+			}
+			// emit the caption, joining the words with a space
+			std::string output;
+			for (const auto &word : emitted) {
+				output += word + " ";
+			}
+			this->callback(output);
+			// push back the words that were emitted, in reverse order
+			for (auto it = emitted.rbegin(); it != emitted.rend(); ++it) {
+				this->wordQueue.push_front(*it);
+			}
+
+			if (this->wordQueue.size() >= this->maxSize ||
+			    std::chrono::steady_clock::now() - startTime >= this->maxTime) {
+				// flush the queue if it's full or we've reached the max time
+				size_t words_to_flush =
+					std::min(this->wordQueue.size(), this->maxSize);
+				for (size_t i = 0; i < words_to_flush; ++i) {
+					wordQueue.pop_front();
+				}
+				startTime = std::chrono::steady_clock::now();
+			}
+
+			newDataAvailable = false;
+		}
+		obs_log(LOG_INFO, "CaptionMonitor::monitor: done");
+	}
+
+	std::deque<std::string> wordQueue;
+	std::thread workerThread;
+	std::mutex queueMutex;
+	std::condition_variable condVar;
+	std::function<void(std::string)> callback;
+	size_t maxSize;
+	std::chrono::seconds maxTime;
+	bool stop;
+	bool initialized = false;
+	bool newDataAvailable = false;
+};
+
+#endif // CAPTIONS_THREAD_H

src/model-utils/model-infos.cpp

@@ -29,7 +29,7 @@ std::map<std::string, ModelInfo> models_info = {{
 	  MODEL_TYPE_TRANSCRIPTION,
 	  {{"https://ggml.ggerganov.com/ggml-model-whisper-base-q5_1.bin",
 	    "422F1AE452ADE6F30A004D7E5C6A43195E4433BC370BF23FAC9CC591F01A8898"}}}},
-	{"Whisper Base En q5 (57Mb)",
+	{"Whisper Base English q5 (57Mb)",
 	 {"Whisper Base En q5",
 	  "ggml-model-whisper-base-en-q5_1",
 	  MODEL_TYPE_TRANSCRIPTION,
@@ -41,7 +41,7 @@ std::map<std::string, ModelInfo> models_info = {{
 	  MODEL_TYPE_TRANSCRIPTION,
 	  {{"https://ggml.ggerganov.com/ggml-model-whisper-base.bin",
 	    "60ED5BC3DD14EEA856493D334349B405782DDCAF0028D4B5DF4088345FBA2EFE"}}}},
-	{"Whisper Base En (141Mb)",
+	{"Whisper Base English (141Mb)",
 	 {"Whisper Base En",
 	  "ggml-model-whisper-base-en",
 	  MODEL_TYPE_TRANSCRIPTION,
@@ -59,7 +59,7 @@ std::map<std::string, ModelInfo> models_info = {{
 	  MODEL_TYPE_TRANSCRIPTION,
 	  {{"https://ggml.ggerganov.com/ggml-model-whisper-medium-q5_0.bin",
 	    "19FEA4B380C3A618EC4723C3EEF2EB785FFBA0D0538CF43F8F235E7B3B34220F"}}}},
-	{"Whisper Medium En q5 (514Mb)",
+	{"Whisper Medium English q5 (514Mb)",
 	 {"Whisper Medium En q5",
 	  "ggml-model-whisper-medium-en-q5_0",
 	  MODEL_TYPE_TRANSCRIPTION,
@@ -71,7 +71,7 @@ std::map<std::string, ModelInfo> models_info = {{
 	  MODEL_TYPE_TRANSCRIPTION,
 	  {{"https://ggml.ggerganov.com/ggml-model-whisper-small-q5_1.bin",
 	    "AE85E4A935D7A567BD102FE55AFC16BB595BDB618E11B2FC7591BC08120411BB"}}}},
-	{"Whisper Small En q5 (181Mb)",
+	{"Whisper Small English q5 (181Mb)",
 	 {"Whisper Small En q5",
 	  "ggml-model-whisper-small-en-q5_1",
 	  MODEL_TYPE_TRANSCRIPTION,
@@ -83,7 +83,7 @@ std::map<std::string, ModelInfo> models_info = {{
 	  MODEL_TYPE_TRANSCRIPTION,
 	  {{"https://ggml.ggerganov.com/ggml-model-whisper-small.bin",
 	    "1BE3A9B2063867B937E64E2EC7483364A79917E157FA98C5D94B5C1FFFEA987B"}}}},
-	{"Whisper Small En (465Mb)",
+	{"Whisper Small English (465Mb)",
 	 {"Whisper Small En",
 	  "ggml-model-whisper-small-en",
 	  MODEL_TYPE_TRANSCRIPTION,
@@ -101,19 +101,19 @@ std::map<std::string, ModelInfo> models_info = {{
 	  MODEL_TYPE_TRANSCRIPTION,
 	  {{"https://ggml.ggerganov.com/ggml-model-whisper-tiny-q5_1.bin",
 	    "818710568DA3CA15689E31A743197B520007872FF9576237BDA97BD1B469C3D7"}}}},
-	{"Whisper Tiny En q5 (31Mb)",
+	{"Whisper Tiny English q5 (31Mb)",
 	 {"Whisper Tiny En q5",
 	  "ggml-model-whisper-tiny-en-q5_1",
 	  MODEL_TYPE_TRANSCRIPTION,
 	  {{"https://ggml.ggerganov.com/ggml-model-whisper-tiny.en-q5_1.bin",
 	    "C77C5766F1CEF09B6B7D47F21B546CBDDD4157886B3B5D6D4F709E91E66C7C2B"}}}},
-	{"Whisper Tiny En q8 (42Mb)",
+	{"Whisper Tiny English q8 (42Mb)",
 	 {"Whisper Tiny En q8",
 	  "ggml-model-whisper-tiny-en-q8_0",
 	  MODEL_TYPE_TRANSCRIPTION,
 	  {{"https://ggml.ggerganov.com/ggml-model-whisper-tiny.en-q8_0.bin",
 	    "5BC2B3860AA151A4C6E7BB095E1FCCE7CF12C7B020CA08DCEC0C6D018BB7DD94"}}}},
-	{"Whisper Tiny En (74Mb)",
+	{"Whisper Tiny English (74Mb)",
 	 {"Whisper Tiny En",
 	  "ggml-model-whisper-tiny-en",
 	  MODEL_TYPE_TRANSCRIPTION,