Fix the decoding issues

CMakeLists.txt

@@ -478,6 +478,7 @@ add_library(${TARGET}
     ${GGML_SOURCES_METAL}
     ${GGML_SOURCES_CUDA}
     ${GGML_SOURCES_OPENCL}
+    unicode.h
     whisper.h
     whisper.cpp
     )

Makefile

@@ -319,7 +319,7 @@ ggml-quants.o: ggml-quants.c ggml.h ggml-quants.h
 
 WHISPER_OBJ += ggml.o ggml-alloc.o ggml-backend.o ggml-quants.o
 
-whisper.o: whisper.cpp whisper.h ggml.h ggml-cuda.h
+whisper.o: whisper.cpp whisper.h unicode.h ggml.h ggml-cuda.h
 	$(CXX) $(CXXFLAGS) -c $< -o $@
 
 ifndef WHISPER_COREML
@@ -358,9 +358,10 @@ CC_SDL=`sdl2-config --cflags --libs`
 
 SRC_COMMON     = examples/common.cpp examples/common-ggml.cpp
 SRC_COMMON_SDL = examples/common-sdl.cpp
+SRC_CONSOLE    = examples/console.cpp
 
 main: examples/main/main.cpp $(SRC_COMMON) $(WHISPER_OBJ)
-	$(CXX) $(CXXFLAGS) examples/main/main.cpp $(SRC_COMMON) $(WHISPER_OBJ) -o main $(LDFLAGS)
+	$(CXX) $(CXXFLAGS) examples/main/main.cpp $(SRC_COMMON) $(SRC_CONSOLE) $(WHISPER_OBJ) -o main $(LDFLAGS)
 	./main -h
 
 bench: examples/bench/bench.cpp $(WHISPER_OBJ)
@@ -373,19 +374,19 @@ server: examples/server/server.cpp $(SRC_COMMON) $(WHISPER_OBJ)
 	$(CXX) $(CXXFLAGS) examples/server/server.cpp $(SRC_COMMON) $(WHISPER_OBJ) -o server $(LDFLAGS) $(LWINSOCK2)
 
 stream: examples/stream/stream.cpp $(SRC_COMMON) $(SRC_COMMON_SDL) $(WHISPER_OBJ)
-	$(CXX) $(CXXFLAGS) examples/stream/stream.cpp $(SRC_COMMON) $(SRC_COMMON_SDL) $(WHISPER_OBJ) -o stream $(CC_SDL) $(LDFLAGS)
+	$(CXX) $(CXXFLAGS) examples/stream/stream.cpp $(SRC_COMMON) $(SRC_CONSOLE) $(SRC_COMMON_SDL) $(WHISPER_OBJ) -o stream $(CC_SDL) $(LDFLAGS)
 
-command: examples/command/command.cpp examples/grammar-parser.cpp $(SRC_COMMON) $(SRC_COMMON_SDL) $(WHISPER_OBJ)
+command: examples/command/command.cpp examples/grammar-parser.cpp $(SRC_COMMON) $(SRC_CONSOLE) $(SRC_COMMON_SDL) $(WHISPER_OBJ)
 	$(CXX) $(CXXFLAGS) examples/command/command.cpp examples/grammar-parser.cpp $(SRC_COMMON) $(SRC_COMMON_SDL) $(WHISPER_OBJ) -o command $(CC_SDL) $(LDFLAGS)
 
 lsp: examples/lsp/lsp.cpp $(SRC_COMMON) $(SRC_COMMON_SDL) $(WHISPER_OBJ)
 	$(CXX) $(CXXFLAGS) examples/lsp/lsp.cpp $(SRC_COMMON) $(SRC_COMMON_SDL) $(WHISPER_OBJ) -o lsp $(CC_SDL) $(LDFLAGS)
 
 talk: examples/talk/talk.cpp examples/talk/gpt-2.cpp $(SRC_COMMON) $(SRC_COMMON_SDL) $(WHISPER_OBJ)
-	$(CXX) $(CXXFLAGS) examples/talk/talk.cpp examples/talk/gpt-2.cpp $(SRC_COMMON) $(SRC_COMMON_SDL) $(WHISPER_OBJ) -o talk $(CC_SDL) $(LDFLAGS)
+	$(CXX) $(CXXFLAGS) examples/talk/talk.cpp examples/talk/gpt-2.cpp $(SRC_COMMON) $(SRC_CONSOLE) $(SRC_COMMON_SDL) $(WHISPER_OBJ) -o talk $(CC_SDL) $(LDFLAGS)
 
 talk-llama: examples/talk-llama/talk-llama.cpp examples/talk-llama/llama.cpp $(SRC_COMMON) $(SRC_COMMON_SDL) $(WHISPER_OBJ)
-	$(CXX) $(CXXFLAGS) examples/talk-llama/talk-llama.cpp examples/talk-llama/llama.cpp $(SRC_COMMON) $(SRC_COMMON_SDL) $(WHISPER_OBJ) -o talk-llama $(CC_SDL) $(LDFLAGS)
+	$(CXX) $(CXXFLAGS) examples/talk-llama/talk-llama.cpp examples/talk-llama/llama.cpp $(SRC_COMMON) $(SRC_CONSOLE) $(SRC_COMMON_SDL) $(WHISPER_OBJ) -o talk-llama $(CC_SDL) $(LDFLAGS)
 
 #
 # Audio samples

bindings/go/params.go

@@ -19,10 +19,6 @@ func (p *Params) SetTranslate(v bool) {
 	p.translate = toBool(v)
 }
 
-func (p *Params) SetSplitOnWord(v bool) {
-	p.split_on_word = toBool(v)
-}
-
 func (p *Params) SetNoContext(v bool) {
 	p.no_context = toBool(v)
 }

bindings/go/pkg/whisper/context.go

@@ -81,10 +81,6 @@ func (context *context) SetSpeedup(v bool) {
 	context.params.SetSpeedup(v)
 }
 
-func (context *context) SetSplitOnWord(v bool) {
-	context.params.SetSplitOnWord(v)
-}
-
 // Set number of threads to use
 func (context *context) SetThreads(v uint) {
 	context.params.SetThreads(int(v))

bindings/go/pkg/whisper/interface.go

@@ -42,7 +42,6 @@ type Context interface {
 	SetDuration(time.Duration)      // Set duration
 	SetThreads(uint)                // Set number of threads to use
 	SetSpeedup(bool)                // Set speedup flag
-	SetSplitOnWord(bool)            // Set split on word flag
 	SetTokenThreshold(float32)      // Set timestamp token probability threshold
 	SetTokenSumThreshold(float32)   // Set timestamp token sum probability threshold
 	SetMaxSegmentLength(uint)       // Set max segment length in characters

bindings/javascript/package.json

@@ -1,26 +1,26 @@
-{
-  "name": "whisper.cpp",
-  "version": "1.5.4",
-  "description": "Whisper speech recognition",
-  "main": "whisper.js",
-  "scripts": {
-    "test": "echo \"todo: add tests\" && exit 0"
-  },
-  "repository": {
-    "type": "git",
-    "url": "git+https://github.com/ggerganov/whisper.cpp"
-  },
-  "keywords": [
-    "openai",
-    "whisper",
-    "speech-to-text",
-    "speech-recognition",
-    "transformer"
-  ],
-  "author": "Georgi Gerganov",
-  "license": "MIT",
-  "bugs": {
-    "url": "https://github.com/ggerganov/whisper.cpp/issues"
-  },
-  "homepage": "https://github.com/ggerganov/whisper.cpp#readme"
-}
+{
+  "name": "whisper.cpp",
+  "version": "1.5.4",
+  "description": "Whisper speech recognition",
+  "main": "whisper.js",
+  "scripts": {
+    "test": "echo \"todo: add tests\" && exit 0"
+  },
+  "repository": {
+    "type": "git",
+    "url": "git+https://github.com/ggerganov/whisper.cpp"
+  },
+  "keywords": [
+    "openai",
+    "whisper",
+    "speech-to-text",
+    "speech-recognition",
+    "transformer"
+  ],
+  "author": "Georgi Gerganov",
+  "license": "MIT",
+  "bugs": {
+    "url": "https://github.com/ggerganov/whisper.cpp/issues"
+  },
+  "homepage": "https://github.com/ggerganov/whisper.cpp#readme"
+}

bindings/ruby/ext/ruby_whisper.cpp

@@ -305,12 +305,6 @@ static VALUE ruby_whisper_params_get_token_timestamps(VALUE self) {
 static VALUE ruby_whisper_params_set_token_timestamps(VALUE self, VALUE value) {
   BOOL_PARAMS_SETTER(self, token_timestamps, value)
 }
-static VALUE ruby_whisper_params_get_split_on_word(VALUE self) {
-  BOOL_PARAMS_GETTER(self, split_on_word)
-}
-static VALUE ruby_whisper_params_set_split_on_word(VALUE self, VALUE value) {
-  BOOL_PARAMS_SETTER(self, split_on_word, value)
-}
 static VALUE ruby_whisper_params_get_speed_up(VALUE self) {
   BOOL_PARAMS_GETTER(self, speed_up)
 }
@@ -406,8 +400,6 @@ void Init_whisper() {
   rb_define_method(cParams, "suppress_non_speech_tokens=", ruby_whisper_params_set_suppress_non_speech_tokens, 1);
   rb_define_method(cParams, "token_timestamps", ruby_whisper_params_get_token_timestamps, 0);
   rb_define_method(cParams, "token_timestamps=", ruby_whisper_params_set_token_timestamps, 1);
-  rb_define_method(cParams, "split_on_word", ruby_whisper_params_get_split_on_word, 0);
-  rb_define_method(cParams, "split_on_word=", ruby_whisper_params_set_split_on_word, 1);
   rb_define_method(cParams, "speed_up", ruby_whisper_params_get_speed_up, 0);
   rb_define_method(cParams, "speed_up=", ruby_whisper_params_set_speed_up, 1);
   rb_define_method(cParams, "diarize", ruby_whisper_params_get_diarize, 0);