Add TTS WebAssembly for NodeJS.

CMakeLists.txt

@@ -23,6 +23,7 @@ option(SHERPA_ONNX_ENABLE_GPU "Enable ONNX Runtime GPU support" OFF)
 option(SHERPA_ONNX_ENABLE_WASM "Whether to enable WASM" OFF)
 option(SHERPA_ONNX_ENABLE_WASM_TTS "Whether to enable WASM for TTS" OFF)
 option(SHERPA_ONNX_ENABLE_WASM_ASR "Whether to enable WASM for ASR" OFF)
+option(SHERPA_ONNX_ENABLE_WASM_NODEJS "Whether to enable WASM for NodeJS" OFF)
 option(SHERPA_ONNX_ENABLE_BINARY "Whether to build binaries" ON)
 option(SHERPA_ONNX_LINK_LIBSTDCPP_STATICALLY "True to link libstdc++ statically. Used only when BUILD_SHARED_LIBS is OFF on Linux" ON)
 
@@ -108,6 +109,7 @@ message(STATUS "SHERPA_ONNX_ENABLE_GPU ${SHERPA_ONNX_ENABLE_GPU}")
 message(STATUS "SHERPA_ONNX_ENABLE_WASM ${SHERPA_ONNX_ENABLE_WASM}")
 message(STATUS "SHERPA_ONNX_ENABLE_WASM_TTS ${SHERPA_ONNX_ENABLE_WASM_TTS}")
 message(STATUS "SHERPA_ONNX_ENABLE_WASM_ASR ${SHERPA_ONNX_ENABLE_WASM_ASR}")
+message(STATUS "SHERPA_ONNX_ENABLE_WASM_NODEJS ${SHERPA_ONNX_ENABLE_WASM_NODEJS}")
 
 if(SHERPA_ONNX_ENABLE_WASM_TTS)
   if(NOT SHERPA_ONNX_ENABLE_WASM)
@@ -121,6 +123,12 @@ if(SHERPA_ONNX_ENABLE_WASM_ASR)
   endif()
 endif()
 
+if(SHERPA_ONNX_ENABLE_WASM_NODEJS)
+  if(NOT SHERPA_ONNX_ENABLE_WASM)
+    message(FATAL_ERROR "Please set SHERPA_ONNX_ENABLE_WASM to ON if you enable WASM for NodeJS")
+  endif()
+endif()
+
 if(SHERPA_ONNX_ENABLE_WASM)
   add_definitions(-DSHERPA_ONNX_ENABLE_WASM=1)
 endif()

build-wasm-simd-nodejs.sh

@@ -0,0 +1,63 @@
+#!/usr/bin/env bash
+# Copyright (c)  2024  Xiaomi Corporation
+#
+# This script is to build sherpa-onnx for WebAssembly (NodeJS)
+#
+# Please use NodeJS >= 18
+
+set -ex
+
+if [ x"$EMSCRIPTEN" == x"" ]; then
+  if ! command -v emcc &> /dev/null; then
+    echo "Please install emscripten first"
+    echo ""
+    echo "You can use the following commands to install it:"
+    echo ""
+    echo "git clone https://github.com/emscripten-core/emsdk.git"
+    echo "cd emsdk"
+    echo "git pull"
+    echo "./emsdk install latest"
+    echo "./emsdk activate latest"
+    echo "source ./emsdk_env.sh"
+    exit 1
+  else
+    EMSCRIPTEN=$(dirname $(realpath $(which emcc)))
+  fi
+fi
+
+export EMSCRIPTEN=$EMSCRIPTEN
+echo "EMSCRIPTEN: $EMSCRIPTEN"
+if [ ! -f $EMSCRIPTEN/cmake/Modules/Platform/Emscripten.cmake ]; then
+  echo "Cannot find $EMSCRIPTEN/cmake/Modules/Platform/Emscripten.cmake"
+  echo "Please make sure you have installed emsdk correctly"
+  exit 1
+fi
+
+mkdir -p build-wasm-simd-nodejs
+pushd build-wasm-simd-nodejs
+
+export SHERPA_ONNX_IS_USING_BUILD_WASM_SH=ON
+
+cmake \
+  -DCMAKE_INSTALL_PREFIX=./install \
+  -DCMAKE_BUILD_TYPE=Release \
+  -DCMAKE_TOOLCHAIN_FILE=$EMSCRIPTEN/cmake/Modules/Platform/Emscripten.cmake \
+  \
+  -DSHERPA_ONNX_ENABLE_PYTHON=OFF \
+  -DSHERPA_ONNX_ENABLE_TESTS=OFF \
+  -DSHERPA_ONNX_ENABLE_CHECK=OFF \
+  -DBUILD_SHARED_LIBS=OFF \
+  -DSHERPA_ONNX_ENABLE_PORTAUDIO=OFF \
+  -DSHERPA_ONNX_ENABLE_JNI=OFF \
+  -DSHERPA_ONNX_ENABLE_C_API=ON \
+  -DSHERPA_ONNX_ENABLE_WEBSOCKET=OFF \
+  -DSHERPA_ONNX_ENABLE_GPU=OFF \
+  -DSHERPA_ONNX_ENABLE_WASM=ON \
+  -DSHERPA_ONNX_ENABLE_WASM_NODEJS=ON \
+  -DSHERPA_ONNX_ENABLE_BINARY=OFF \
+  -DSHERPA_ONNX_LINK_LIBSTDCPP_STATICALLY=OFF \
+  ..
+make -j10
+make install
+
+ls -lh install/bin/wasm/nodejs

wasm/CMakeLists.txt

@@ -5,3 +5,7 @@ endif()
 if(SHERPA_ONNX_ENABLE_WASM_ASR)
   add_subdirectory(asr)
 endif()
+
+if(SHERPA_ONNX_ENABLE_WASM_NODEJS)
+  add_subdirectory(nodejs)
+endif()

wasm/asr/sherpa-onnx-wasm-main-asr.cc

@@ -1,4 +1,4 @@
-// wasm/sherpa-onnx-wasm-asr-main.cc
+// wasm/sherpa-onnx-wasm-main-asr.cc
 //
 // Copyright (c)  2024  Xiaomi Corporation
 #include <stdio.h>

wasm/nodejs/CMakeLists.txt

@@ -0,0 +1,66 @@
+if(NOT $ENV{SHERPA_ONNX_IS_USING_BUILD_WASM_SH})
+  message(FATAL_ERROR "Please use ./build-wasm-simd-nodejs.sh to build for wasm NodeJS")
+endif()
+
+set(exported_functions
+  MyPrintOfflineTtsConfig
+  MyPrint
+  #tts
+  SherpaOnnxCreateOfflineTts
+  SherpaOnnxDestroyOfflineTts
+  SherpaOnnxDestroyOfflineTtsGeneratedAudio
+  SherpaOnnxOfflineTtsGenerate
+  SherpaOnnxOfflineTtsGenerateWithCallback
+  SherpaOnnxOfflineTtsNumSpeakers
+  SherpaOnnxOfflineTtsSampleRate
+  SherpaOnnxWriteWave
+  # streaming asr
+  AcceptWaveform
+  CreateOnlineRecognizer
+  CreateOnlineStream
+  DecodeOnlineStream
+  DestroyOnlineRecognizer
+  DestroyOnlineRecognizerResult
+  DestroyOnlineStream
+  GetOnlineStreamResult
+  InputFinished
+  IsEndpoint
+  IsOnlineStreamReady
+  Reset
+  # non-streaming ASR
+)
+
+set(mangled_exported_functions)
+foreach(x IN LISTS exported_functions)
+  list(APPEND mangled_exported_functions "_${x}")
+endforeach()
+list(JOIN mangled_exported_functions "," all_exported_functions)
+
+include_directories(${CMAKE_SOURCE_DIR})
+set(MY_FLAGS " -s FORCE_FILESYSTEM=1 -s INITIAL_MEMORY=512MB -s ALLOW_MEMORY_GROWTH=1")
+string(APPEND MY_FLAGS " -sSTACK_SIZE=10485760 ") # 10MB
+string(APPEND MY_FLAGS " -sEXPORTED_FUNCTIONS=[_CopyHeap,_malloc,_free,${all_exported_functions}] ")
+string(APPEND MY_FLAGS " -sEXPORTED_RUNTIME_METHODS=['ccall','stringToUTF8','setValue','getValue'] ")
+string(APPEND MY_FLAGS " -sNODERAWFS=1 ")
+string(APPEND MY_FLAGS " -sEXPORTED_RUNTIME_METHODS=['ccall','stringToUTF8','setValue','getValue','lengthBytesUTF8','UTF8ToString'] ")
+string(APPEND MY_FLAGS " -sMODULARIZE=1 -sWASM_ASYNC_COMPILATION=0 ")
+
+message(STATUS "MY_FLAGS: ${MY_FLAGS}")
+
+set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${MY_FLAGS}")
+set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${MY_FLAGS}")
+set(CMAKE_EXECUTBLE_LINKER_FLAGS "${CMAKE_EXECUTBLE_LINKER_FLAGS} ${MY_FLAGS}")
+
+add_executable(sherpa-onnx-wasm-nodejs sherpa-onnx-wasm-nodejs.cc)
+target_link_libraries(sherpa-onnx-wasm-nodejs sherpa-onnx-core sherpa-onnx-c-api)
+install(TARGETS sherpa-onnx-wasm-nodejs DESTINATION bin/wasm/nodejs)
+
+install(
+  FILES
+  ${CMAKE_SOURCE_DIR}/wasm/asr/sherpa-onnx-asr.js
+  ${CMAKE_SOURCE_DIR}/wasm/tts/sherpa-onnx-tts.js
+    "$<TARGET_FILE_DIR:sherpa-onnx-wasm-nodejs>/sherpa-onnx-wasm-nodejs.js"
+    "$<TARGET_FILE_DIR:sherpa-onnx-wasm-nodejs>/sherpa-onnx-wasm-nodejs.wasm"
+  DESTINATION
+    bin/wasm/nodejs
+)

wasm/nodejs/sherpa-onnx-wasm-nodejs.cc

@@ -0,0 +1,40 @@
+// wasm/sherpa-onnx-wasm-main-nodejs.cc
+//
+// Copyright (c)  2024  Xiaomi Corporation
+#include <stdio.h>
+
+#include <algorithm>
+#include <memory>
+
+#include "sherpa-onnx/c-api/c-api.h"
+
+extern "C" {
+
+void MyPrintOfflineTtsConfig(SherpaOnnxOfflineTtsConfig *tts_config) {
+  auto tts_model_config = &tts_config->model;
+  auto vits_model_config = &tts_model_config->vits;
+  fprintf(stdout, "----------vits model config----------\n");
+  fprintf(stdout, "model: %s\n", vits_model_config->model);
+  fprintf(stdout, "lexicon: %s\n", vits_model_config->lexicon);
+  fprintf(stdout, "tokens: %s\n", vits_model_config->tokens);
+  fprintf(stdout, "data_dir: %s\n", vits_model_config->data_dir);
+  fprintf(stdout, "noise scale: %.3f\n", vits_model_config->noise_scale);
+  fprintf(stdout, "noise scale w: %.3f\n", vits_model_config->noise_scale_w);
+  fprintf(stdout, "length scale: %.3f\n", vits_model_config->length_scale);
+
+  fprintf(stdout, "----------tts model config----------\n");
+  fprintf(stdout, "num threads: %d\n", tts_model_config->num_threads);
+  fprintf(stdout, "debug: %d\n", tts_model_config->debug);
+  fprintf(stdout, "provider: %s\n", tts_model_config->provider);
+
+  fprintf(stdout, "----------tts config----------\n");
+  fprintf(stdout, "rule_fsts: %s\n", tts_config->rule_fsts);
+  fprintf(stdout, "max num sentences: %d\n", tts_config->max_num_sentences);
+}
+
+void MyPrint(SherpaOnnxOnlineRecognizerConfig *config) {}
+
+void CopyHeap(const char *src, int32_t num_bytes, char *dst) {
+  std::copy(src, src + num_bytes, dst);
+}
+}

wasm/tts/app-tts.js

@@ -22,7 +22,7 @@ Module.onRuntimeInitialized = function() {
   console.log('Model files downloaded!');
 
   console.log('Initializing tts ......');
-  tts = initSherpaOnnxOfflineTts()
+  tts = initSherpaOnnxOfflineTts(Module)
   if (tts.numSpeakers > 1) {
     speakerIdLabel.innerHTML = `Speaker ID (0 - ${tts.numSpeakers - 1}):`;
   }