[http] Update http server (#189)

Co-authored-by: 李盛强 <lichengqiang@lichengqiangdeMacBook-Pro.local>

runtime/core/bin/http_server_main.cc

@@ -16,36 +16,69 @@
 #include "glog/logging.h"
 
 #include "http/http_server.h"
-#include "utils/log.h"
+#include "processor/wetext_processor.h"
 
+#include "frontend/g2p_en.h"
+#include "frontend/g2p_prosody.h"
+#include "frontend/wav.h"
+#include "model/tts_model.h"
+#include "utils/string.h"
+
+// Flags
+DEFINE_string(frontend_flags, "", "frontend flags file");
+DEFINE_string(vits_flags, "", "vits flags file");
+
+// Text Normalization
 DEFINE_string(tagger, "", "tagger fst file");
 DEFINE_string(verbalizer, "", "verbalizer fst file");
 
+// Tokenizer
 DEFINE_string(vocab, "", "tokenizer vocab file");
 
+// G2P for English
+DEFINE_string(cmudict, "", "cmudict for english words");
+DEFINE_string(g2p_en_model, "", "english g2p fst model for oov");
+DEFINE_string(g2p_en_sym, "", "english g2p symbol table for oov");
+
+// G2P for Chinese
 DEFINE_string(char2pinyin, "", "chinese character to pinyin");
 DEFINE_string(pinyin2id, "", "pinyin to id");
 DEFINE_string(pinyin2phones, "", "pinyin to phones");
 DEFINE_string(g2p_prosody_model, "", "g2p prosody model file");
 
+// VITS
 DEFINE_string(speaker2id, "", "speaker to id");
 DEFINE_string(phone2id, "", "phone to id");
 DEFINE_string(vits_model, "", "e2e tts model file");
+DEFINE_int32(sampling_rate, 22050, "sampling rate of pcm");
 
+
+// port
 DEFINE_int32(port, 10086, "http listening port");
 
 int main(int argc, char* argv[]) {
   gflags::ParseCommandLineFlags(&argc, &argv, false);
   google::InitGoogleLogging(argv[0]);
+  gflags::ReadFromFlagsFile(FLAGS_frontend_flags, "", false);
+  gflags::ReadFromFlagsFile(FLAGS_vits_flags, "", false);
 
   auto tn = std::make_shared<wetext::Processor>(FLAGS_tagger, FLAGS_verbalizer);
+
+  bool has_en = !FLAGS_g2p_en_model.empty() && !FLAGS_g2p_en_sym.empty() &&
+                !FLAGS_g2p_en_sym.empty();
+  std::shared_ptr<wetts::G2pEn> g2p_en =
+      has_en ? std::make_shared<wetts::G2pEn>(FLAGS_cmudict, FLAGS_g2p_en_model,
+                                              FLAGS_g2p_en_sym)
+             : nullptr;
+
   auto g2p_prosody = std::make_shared<wetts::G2pProsody>(
       FLAGS_g2p_prosody_model, FLAGS_vocab, FLAGS_char2pinyin, FLAGS_pinyin2id,
-      FLAGS_pinyin2phones);
-  auto tts_model = std::make_shared<wetts::TtsModel>(
-      FLAGS_vits_model, FLAGS_speaker2id, FLAGS_phone2id, tn, g2p_prosody);
+      FLAGS_pinyin2phones, g2p_en);
+  auto model = std::make_shared<wetts::TtsModel>(
+      FLAGS_vits_model, FLAGS_speaker2id, FLAGS_phone2id, FLAGS_sampling_rate,
+      tn, g2p_prosody);
 
-  wetts::HttpServer server(FLAGS_port, tts_model);
+  wetts::HttpServer server(FLAGS_port, model);
   LOG(INFO) << "Listening at port " << FLAGS_port;
   server.Start();
   return 0;

runtime/core/bin/tts_main.cc

@@ -73,7 +73,8 @@ int main(int argc, char* argv[]) {
       FLAGS_g2p_prosody_model, FLAGS_vocab, FLAGS_char2pinyin, FLAGS_pinyin2id,
       FLAGS_pinyin2phones, g2p_en);
   auto model = std::make_shared<wetts::TtsModel>(
-      FLAGS_vits_model, FLAGS_speaker2id, FLAGS_phone2id, tn, g2p_prosody);
+      FLAGS_vits_model, FLAGS_speaker2id, FLAGS_phone2id, FLAGS_sampling_rate,
+      tn, g2p_prosody);
 
   std::vector<float> audio;
   int sid = model->GetSid(FLAGS_sname);

runtime/core/http/http_server.cc

@@ -25,29 +25,25 @@
 
 #include "frontend/wav.h"
 #include "utils/string.h"
+#include "utils/timer.h"
 
 namespace wetts {
 
 namespace urls = boost::urls;
 namespace uuids = boost::uuids;
 
-http::message_generator ConnectionHandler::handle_request(
-    const std::string& wav_path) {
-  // Attempt to open the file
+http::message_generator ConnectionHandler::HandleRequest(
+  char* wav_data, int data_size) {
   beast::error_code ec;
-  http::file_body::value_type body;
-  body.open(wav_path.c_str(), beast::file_mode::scan, ec);
-
-  // Cache the size since we need it after the move
-  auto const size = body.size();
-  // Respond to GET request
-  http::response<http::file_body> res{
-      std::piecewise_construct, std::make_tuple(std::move(body)),
-      std::make_tuple(http::status::ok, request_.version())};
+  http::response<http::buffer_body> res;
+  res.result(http::status::ok);
   res.set(http::field::server, BOOST_BEAST_VERSION_STRING);
   res.set(http::field::content_type, "audio/wav");
-  res.content_length(size);
   res.keep_alive(request_.keep_alive());
+  res.body().data = wav_data;
+  res.body().size = data_size;
+  res.body().more = false;
+  res.prepare_payload();
   return res;
 }
 
@@ -80,16 +76,37 @@ void ConnectionHandler::operator()() {
     std::string name = (*params.find("name")).value;
     int sid = tts_model_->GetSid(name);
     // 2. Synthesis audio from text
-    std::vector<float> audio;
-    tts_model_->Synthesis(text, sid, &audio);
-    wetts::WavWriter wav_writer(audio.data(), audio.size(), 1, 22050, 16);
-    // 3. Write samples to file named uuid.wav
-    std::string wav_path =
-        uuids::to_string(uuids::random_generator()()) + ".wav";
-    wav_writer.Write(wav_path);
-
+    int sample_rate = tts_model_->sampling_rate();
+    int num_channels = 1;
+    int bits_per_sample = 16;
+    LOG(INFO) << "Sample rate: " << sample_rate;
+    LOG(INFO) << "Num of channels: " << num_channels;
+    LOG(INFO) << "Bit per sample: " << bits_per_sample;
+    int extract_time = 0;
+    wetts::Timer timer;
+    std::vector<float> pcm;
+    tts_model_->Synthesis(text, sid, &pcm);
+    int pcm_size = pcm.size();
+    extract_time = timer.Elapsed();
+    LOG(INFO) << "TTS pcm duration: "
+              << pcm_size * 1000 / num_channels / sample_rate << "ms";
+    LOG(INFO) << "Cost time: " << static_cast<float>(extract_time) << "ms";
+    // 3. Convert pcm to wav
+    std::vector<int16_t> audio(pcm_size);
+    for (int i = 0; i < pcm_size; ++i) {
+      audio[i] = static_cast<int16_t>(pcm[i]);
+    }
+    int audio_size = pcm_size * sizeof(int16_t);
+    int data_size = audio_size + 44;
+    WavHeader header(pcm_size, num_channels, sample_rate, bits_per_sample);
+    std::vector<char> wav_data;
+    wav_data.insert(wav_data.end(), reinterpret_cast<char*>(&header),
+                    reinterpret_cast<char*>(&header) + 44);
+    wav_data.insert(wav_data.end(), reinterpret_cast<char*>(audio.data()),
+                    reinterpret_cast<char*>(audio.data()) + audio_size);
     // Handle request
-    http::message_generator msg = handle_request(wav_path);
+    http::message_generator msg =
+        HandleRequest(wav_data.data(), data_size);
     // Determine if we should close the connection
     bool keep_alive = msg.keep_alive();
     // Send the response

runtime/core/http/http_server.h

@@ -38,7 +38,7 @@ class ConnectionHandler {
   ConnectionHandler(tcp::socket&& socket, std::shared_ptr<TtsModel> tts_model)
       : socket_(std::move(socket)), tts_model_(std::move(tts_model)) {}
   void operator()();
-  http::message_generator handle_request(const std::string& wav_path);
+  http::message_generator HandleRequest(char* wav_data, int data_size);
 
  private:
   tcp::socket socket_;

runtime/core/model/tts_model.cc

@@ -26,12 +26,13 @@
 namespace wetts {
 
 TtsModel::TtsModel(const std::string& model_path, const std::string& speaker2id,
-                   const std::string& phone2id,
+                   const std::string& phone2id, const int sampling_rate,
                    std::shared_ptr<wetext::Processor> tn,
                    std::shared_ptr<G2pProsody> g2p_prosody)
     : OnnxModel(model_path),
       tn_(std::move(tn)),
       g2p_prosody_(std::move(g2p_prosody)) {
+  sampling_rate_ = sampling_rate;
   ReadTableFile(phone2id, &phone2id_);
   ReadTableFile(speaker2id, &speaker2id_);
 }

runtime/core/model/tts_model.h

@@ -30,14 +30,17 @@ namespace wetts {
 class TtsModel : public OnnxModel {
  public:
   explicit TtsModel(const std::string& model_path,
-                    const std::string& speaker2id, const std::string& phone2id,
+                    const std::string& speaker2id,
+                    const std::string& phone2id,
+                    const int sampling_rate,
                     std::shared_ptr<wetext::Processor> processor,
                     std::shared_ptr<G2pProsody> g2p_prosody);
   void Forward(const std::vector<int64_t>& phonemes, const int sid,
                std::vector<float>* audio);
   void Synthesis(const std::string& text, const int sid,
                  std::vector<float>* audio);
   int GetSid(const std::string& name);
+  int sampling_rate() const { return sampling_rate_; }
 
  private:
   int sampling_rate_;

runtime/core/utils/timer.h

@@ -0,0 +1,39 @@
+// Copyright (c) 2021 Mobvoi Inc (Binbin Zhang)
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//   http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#ifndef UTILS_TIMER_H_
+#define UTILS_TIMER_H_
+
+#include <chrono>
+
+namespace wetts {
+
+class Timer {
+ public:
+  Timer() : time_start_(std::chrono::steady_clock::now()) {}
+  void Reset() { time_start_ = std::chrono::steady_clock::now(); }
+  // return int in milliseconds
+  int Elapsed() const {
+    auto time_now = std::chrono::steady_clock::now();
+    return std::chrono::duration_cast<std::chrono::milliseconds>(time_now -
+                                                                 time_start_)
+        .count();
+  }
+
+ private:
+  std::chrono::time_point<std::chrono::steady_clock> time_start_;
+};
+}  // namespace wetts
+
+#endif  // UTILS_TIMER_H_