[update] update llm-tts supported kokoro tts

Author: LittleMouse

projects/llm_framework/SConstruct

@@ -5,7 +5,7 @@ import shutil
 os.environ['SDK_PATH'] = os.path.normpath(str(Path(os.getcwd())/'..'/'..'/'SDK'))
 os.environ['EXT_COMPONENTS_PATH'] = os.path.normpath(str(Path(os.getcwd())/'..'/'..'/'ext_components'))
 
-version = 'v0.1.3'
+version = 'v0.1.5'
 static_lib = 'static_lib'
 update = False
 

projects/llm_framework/main_tts/SConstruct

@@ -8,7 +8,7 @@ with open(env['PROJECT_TOOL_S']) as f:
 SRCS = Glob('src/*.c*')
 INCLUDE = [ADir('include'), ADir('.')]
 PRIVATE_INCLUDE = []
-REQUIREMENTS = ['pthread', 'utilities', 'gomp', 'eventpp', 'StackFlow']
+REQUIREMENTS = ['pthread', 'utilities', 'ax_msp', 'gomp', 'eventpp', 'StackFlow']
 STATIC_LIB = []
 DYNAMIC_LIB = []
 DEFINITIONS = []
@@ -20,14 +20,18 @@ STATIC_FILES = []
 DEFINITIONS += ['-O3', '-fopenmp', '-std=c++17']
 LDFLAGS+=['-Wl,-rpath=/opt/m5stack/lib', '-Wl,-rpath=/usr/local/m5stack/lib', '-Wl,-rpath=/usr/local/m5stack/lib/gcc-10.3', '-Wl,-rpath=/opt/lib', '-Wl,-rpath=/opt/usr/lib', '-Wl,-rpath=./']
 LINK_SEARCH_PATH += [ADir('../static_lib')]
-REQUIREMENTS += ['tts']
+REQUIREMENTS += ['ax_engine', 'ax_interpreter', 'ax_sys']
+REQUIREMENTS += ['samplerate']
+REQUIREMENTS += ['tts', 'kokoro', 'libkokoro_backend']
 
 
 INCLUDE += [ADir('src/runner/eigen-3.4.0'), ADir('src/runner/src/tn/header'), ADir('src/runner/include'), ADir('src/runner/src/header')]
 
+REQUIREMENTS += ['onnxruntime']
+
 STATIC_FILES += Glob('mode_*.json')
 
-env['COMPONENTS'].append({'target':'llm_tts-1.6',
+env['COMPONENTS'].append({'target':'llm_tts-1.7',
                           'SRCS':SRCS,
                           'INCLUDE':INCLUDE,
                           'PRIVATE_INCLUDE':PRIVATE_INCLUDE,

projects/llm_framework/main_tts/mode_kokoro-ax650.json

@@ -0,0 +1,33 @@
+{
+    "mode": "kokoro-ax650",
+    "type": "tts",
+    "homepage": "https://huggingface.co/AXERA-TECH/kokoro.axera",
+    "capabilities": [
+        "tts",
+        "Chinese",
+        "English"
+    ],
+    "input_type": [
+        "tts.utf-8"
+    ],
+    "output_type": [
+        "tts.wav",
+        "sys.play.0_1"
+    ],
+    "mode_param": {
+        "axmodel_dir": "./models",
+        "lang": "z",
+        "voice_path": "./voices",
+        "voice_name": "zf_xiaoxiao",
+        "vocab_path": "./vocab.txt",
+        "espeak_data_path": "./espeak-ng-data",
+        "dict_dir": "./dict",
+        "spacker_speed": 1.0,
+        "max_len": 96,
+        "mode_rate": 24000,
+        "audio_rate": 48000,
+        "speed": 1.0,
+        "pause": 0.0,
+        "fade_out": 0.0
+    }
+}

projects/llm_framework/main_tts/src/Kokoro.h

@@ -0,0 +1,68 @@
+/*
+ * SPDX-FileCopyrightText: 2024 M5Stack Technology CO LTD
+ *
+ * SPDX-License-Identifier: MIT
+ */
+#pragma once
+
+#include <string>
+#include <vector>
+#include <memory>
+
+namespace kokoro {
+struct SentenceInfo {
+    std::string sentence;
+    std::vector<int> input_ids;
+    std::string phonemes;
+    int content_len = 0;
+    bool is_long    = false;
+    std::vector<SentenceInfo> sub_results;
+
+    SentenceInfo() = default;
+    SentenceInfo(const std::string &s, const std::vector<int> &ids, const std::string &ph, int len)
+        : sentence(s), input_ids(ids), phonemes(ph), content_len(len), is_long(false)
+    {
+    }
+
+    SentenceInfo(const std::string &s, const std::vector<SentenceInfo> &subs)
+        : sentence(s), is_long(true), sub_results(subs)
+    {
+    }
+};
+
+struct MergedGroup {
+    bool is_long_split = false;
+    std::vector<int> input_ids;
+    std::string phonemes;
+    std::vector<SentenceInfo> sub_results;
+
+    MergedGroup() = default;
+    MergedGroup(const std::vector<int> &ids, const std::string &ph) : is_long_split(false), input_ids(ids), phonemes(ph)
+    {
+    }
+
+    MergedGroup(const std::vector<SentenceInfo> &subs) : is_long_split(true), sub_results(subs)
+    {
+    }
+};
+
+class Kokoro {
+public:
+    Kokoro();
+    ~Kokoro();
+    Kokoro(const Kokoro &)            = delete;
+    Kokoro &operator=(const Kokoro &) = delete;
+
+    bool init(const std::string &model_path, int max_seq_len = 96, const std::string &lang_code = "z",
+              const std::string &voices_path = "./voices", const std::string &voice_name = "af_heart",
+              const std::string &vocab_path       = "dict/vocab.txt",
+              const std::string &espeak_data_path = "./espeak-ng-data", const std::string &dict_dir = "./dict");
+
+    bool tts(const std::string &text, const std::string &voice_name, float speed, int sample_rate, float fade_out,
+             float pause_duration, std::vector<float> &generated_audio);
+
+private:
+    struct Impl;
+    std::unique_ptr<Impl> impl_;
+};
+}  // namespace kokoro