diff --git a/src/M5ModuleLLM.cpp b/src/M5ModuleLLM.cpp
index deb8426..b7f4699 100644
--- a/src/M5ModuleLLM.cpp
+++ b/src/M5ModuleLLM.cpp
@@ -14,8 +14,10 @@ bool M5ModuleLLM::begin(Stream* serialPort)
     llm.init(&msg);
     audio.init(&msg);
     tts.init(&msg);
+    melotts.init(&msg);
     kws.init(&msg);
     asr.init(&msg);
+    yolo.init(&msg);
     return true;
 }
 
diff --git a/src/M5ModuleLLM.h b/src/M5ModuleLLM.h
index 31f2a72..eaed899 100644
--- a/src/M5ModuleLLM.h
+++ b/src/M5ModuleLLM.h
@@ -14,6 +14,7 @@
 #include "api/api_melotts.h"
 #include "api/api_kws.h"
 #include "api/api_asr.h"
+#include "api/api_yolo.h"
 
 class M5ModuleLLM {
 public:
@@ -82,6 +83,12 @@ public:
      */
     m5_module_llm::ApiAsr asr;
 
+    /**
+     * @brief YOLO module api set
+     *
+     */
+    m5_module_llm::ApiYolo yolo;
+
     /**
      * @brief MSG module to handle module response message
      *
@@ -100,8 +107,10 @@ private:
 typedef std::function<void(void)> OnKeywordDetectedCallback_t;
 typedef std::function<void(String data, bool isFinish, int index)> OnAsrDataInputCallback_t;
 typedef std::function<void(String data, bool isFinish, int index)> OnLlmDataInputCallback_t;
+typedef std::function<void(String data, bool isFinish, int index)> OnYoloDataInputCallback_t;
 typedef std::function<void(String rawData)> OnAsrDataInputRawCallback_t;
 typedef std::function<void(String rawData)> OnLlmDataInputRawCallback_t;
+typedef std::function<void(String rawData)> OnYoloDataInputRawCallback_t;
 
 /**
  * @brief Voice assistant preset base on class M5ModuleLLM
diff --git a/src/api/api_melotts.cpp b/src/api/api_melotts.cpp
index 9fafe4b..e1cd73f 100644
--- a/src/api/api_melotts.cpp
+++ b/src/api/api_melotts.cpp
@@ -39,7 +39,7 @@ String ApiMelotts::setup(ApiMelottsSetupConfig_t config, String request_id)
             // Copy work id
             work_id = msg.work_id;
         },
-        10000);
+        15000);
     return work_id;
 }
 
@@ -51,7 +51,7 @@ int ApiMelotts::inference(String work_id, String input, uint32_t timeout, String
         doc["request_id"]     = request_id;
         doc["work_id"]        = work_id;
         doc["action"]         = "inference";
-        doc["object"]         = "melotts.utf-8.stream";
+        doc["object"]         = "tts.utf-8.stream";
         doc["data"]["delta"]  = input;
         doc["data"]["index"]  = 0;
         doc["data"]["finish"] = true;
diff --git a/src/api/api_yolo.cpp b/src/api/api_yolo.cpp
new file mode 100644
index 0000000..2774cc0
--- /dev/null
+++ b/src/api/api_yolo.cpp
@@ -0,0 +1,106 @@
+/*
+ * SPDX-FileCopyrightText: 2024 M5Stack Technology CO LTD
+ *
+ * SPDX-License-Identifier: MIT
+ */
+#include "api_yolo.h"
+
+using namespace m5_module_llm;
+
+void ApiYolo::init(ModuleMsg* moduleMsg)
+{
+    _module_msg = moduleMsg;
+}
+
+String ApiYolo::setup(ApiYoloSetupConfig_t config, String request_id)
+{
+    String cmd;
+    {
+        JsonDocument doc;
+        doc["request_id"]              = request_id;
+        doc["work_id"]                 = "yolo";
+        doc["action"]                  = "setup";
+        doc["object"]                  = "yolo.setup";
+        doc["data"]["model"]           = config.model;
+        doc["data"]["response_format"] = config.response_format;
+        JsonArray inputArray           = doc["data"]["input"].to<JsonArray>();
+        for (const String& str : config.input) {
+            inputArray.add(str);
+        }
+        doc["data"]["enoutput"] = config.enoutput;
+        serializeJson(doc, cmd);
+    }
+
+    String work_id;
+    _module_msg->sendCmdAndWaitToTakeMsg(
+        cmd.c_str(), request_id,
+        [&work_id](ResponseMsg_t& msg) {
+            // Copy work id
+            work_id = msg.work_id;
+        },
+        5000);
+    return work_id;
+}
+
+int ApiYolo::inference(String& work_id, uint8_t* input, size_t& raw_len, String request_id)
+{
+    String cmd;
+    {
+        JsonDocument doc;
+        doc["RAW"]        = raw_len;
+        doc["request_id"] = request_id;
+        doc["work_id"]    = work_id;
+        doc["action"]     = "inference";
+        doc["object"]     = "cv.jpeg.base64";
+        serializeJson(doc, cmd);
+    }
+
+    _module_msg->sendCmd(cmd.c_str());
+    _module_msg->sendRaw(input, raw_len);
+    return MODULE_LLM_OK;
+}
+
+int ApiYolo::inferenceAndWaitResult(String& work_id, uint8_t* input, size_t& raw_len,
+                                    std::function<void(String&)> onResult, uint32_t timeout, String request_id)
+{
+    inference(work_id, input, raw_len, request_id);
+
+    uint32_t time_out_count = millis();
+    bool is_time_out        = false;
+    bool is_msg_finish      = false;
+    while (1) {
+        _module_msg->update();
+        _module_msg->takeMsg(request_id, [&time_out_count, &is_msg_finish, &onResult](ResponseMsg_t& msg) {
+            String response_msg;
+            {
+                JsonDocument doc;
+                deserializeJson(doc, msg.raw_msg);
+                response_msg = doc["data"]["delta"].as<String>();
+                if (!doc["data"]["finish"].isNull()) {
+                    is_msg_finish = doc["data"]["finish"];
+                    if (is_msg_finish) {
+                        response_msg += '\n';
+                    }
+                }
+            }
+            if (onResult) {
+                onResult(response_msg);
+            }
+            time_out_count = millis();
+        });
+
+        if (is_msg_finish) {
+            break;
+        }
+
+        if (millis() - time_out_count > timeout) {
+            is_time_out = true;
+            break;
+        }
+    }
+
+    if (is_time_out) {
+        return MODULE_LLM_WAIT_RESPONSE_TIMEOUT;
+    }
+    return MODULE_LLM_OK;
+}
diff --git a/src/api/api_yolo.h b/src/api/api_yolo.h
new file mode 100644
index 0000000..0817aa4
--- /dev/null
+++ b/src/api/api_yolo.h
@@ -0,0 +1,59 @@
+/*
+ * SPDX-FileCopyrightText: 2024 M5Stack Technology CO LTD
+ *
+ * SPDX-License-Identifier: MIT
+ */
+#pragma once
+#include "../utils/msg.h"
+#include <Arduino.h>
+
+namespace m5_module_llm {
+struct ApiYoloSetupConfig_t {
+    String model              = "yolo11n";
+    String response_format    = "yolo.yolobox.stream";
+    std::vector<String> input = {"yolo.jpeg.base64"};
+    bool enoutput             = true;
+};
+
+class ApiYolo {
+public:
+    void init(ModuleMsg* moduleMsg);
+
+    /**
+     * @brief Setup module YOLO, return YOLO work_id
+     *
+     * @param config
+     * @param request_id
+     * @return String
+     */
+    String setup(ApiYoloSetupConfig_t config = ApiYoloSetupConfig_t(), String request_id = "yolo_setup");
+
+    /**
+     * @brief Inference input data by module LLM
+     *
+     * @param raw_len
+     * @param work_id
+     * @param input
+     * @param request_id
+     * @return int
+     */
+    int inference(String& work_id, uint8_t* input, size_t& raw_len, String request_id = "yolo_inference");
+
+    /**
+     * @brief Inference input data by module LLM, and wait inference result
+     *
+     * @param raw_len
+     * @param work_id
+     * @param input
+     * @param onResult On inference result callback
+     * @param timeout
+     * @param request_id
+     * @return int
+     */
+    int inferenceAndWaitResult(String& work_id, uint8_t* input, size_t& raw_len, std::function<void(String&)> onResult,
+                               uint32_t timeout = 5000, String request_id = "yolo_inference");
+
+private:
+    ModuleMsg* _module_msg = nullptr;
+};
+}  // namespace m5_module_llm
diff --git a/src/utils/comm.cpp b/src/utils/comm.cpp
index e835b1a..7df0c4a 100644
--- a/src/utils/comm.cpp
+++ b/src/utils/comm.cpp
@@ -26,6 +26,11 @@ void ModuleComm::sendCmd(const char* cmd)
     _serial->print(cmd);
 }
 
+void ModuleComm::sendRaw(const uint8_t* data, size_t& raw_len)
+{
+    _serial->write(data, raw_len);
+}
+
 ModuleComm::Respond_t ModuleComm::getResponse(uint32_t timeout)
 {
     Respond_t ret;
diff --git a/src/utils/comm.h b/src/utils/comm.h
index fa616d1..99c0344 100644
--- a/src/utils/comm.h
+++ b/src/utils/comm.h
@@ -24,6 +24,7 @@ public:
 
     bool init(Stream* serialPort);
     void sendCmd(const char* cmd);
+    void sendRaw(const uint8_t* data, size_t& raw_len);
     Respond_t getResponse(uint32_t timeout = 0xFFFFFFFF);
 
 private:
diff --git a/src/utils/msg.h b/src/utils/msg.h
index b73e414..60ad69d 100644
--- a/src/utils/msg.h
+++ b/src/utils/msg.h
@@ -59,6 +59,10 @@ public:
         _module_comm->sendCmd(cmd);
     }
 
+    inline void sendRaw(const uint8_t* data, size_t& raw_len)
+    {
+        _module_comm->sendRaw(data, raw_len);
+    }
     /**
      * @brief Module response message list
      *