diff --git a/metagpt/learn/text_to_embedding.py b/metagpt/learn/text_to_embedding.py
index 6d0cefcdb..5c08ef0b9 100644
--- a/metagpt/learn/text_to_embedding.py
+++ b/metagpt/learn/text_to_embedding.py
@@ -16,7 +16,7 @@ from metagpt.utils.common import initialize_environment
 @skill_metadata(name="Text to Embedding",
                 description="Convert the text into embeddings.",
                 requisite="`OPENAI_API_KEY`")
-def text_to_embedding(text, model="text-embedding-ada-002", openai_api_key="", **kwargs):
+async def text_to_embedding(text, model="text-embedding-ada-002", openai_api_key="", **kwargs):
     """Text to embedding
 
     :param text: The text used for embedding.
@@ -26,5 +26,5 @@ def text_to_embedding(text, model="text-embedding-ada-002", openai_api_key="", *
     """
     initialize_environment()
     if os.environ.get("OPENAI_API_KEY") or openai_api_key:
-        return oas3_openai_text_to_embedding(text, model=model, openai_api_key=openai_api_key)
+        return await oas3_openai_text_to_embedding(text, model=model, openai_api_key=openai_api_key)
     raise EnvironmentError
diff --git a/metagpt/learn/text_to_image.py b/metagpt/learn/text_to_image.py
index d245b06db..db9844c71 100644
--- a/metagpt/learn/text_to_image.py
+++ b/metagpt/learn/text_to_image.py
@@ -17,7 +17,7 @@ from metagpt.utils.common import initialize_environment
 @skill_metadata(name="Text to image",
                 description="Create a drawing based on the text.",
                 requisite="`OPENAI_API_KEY` or `METAGPT_TEXT_TO_IMAGE_MODEL`")
-def text_to_image(text, size_type: str = "512x512", openai_api_key="", model_url="", **kwargs):
+async def text_to_image(text, size_type: str = "512x512", openai_api_key="", model_url="", **kwargs):
     """Text to image
 
     :param text: The text used for image conversion.
@@ -29,10 +29,11 @@ def text_to_image(text, size_type: str = "512x512", openai_api_key="", model_url
     initialize_environment()
     image_declaration = "data:image/png;base64,"
     if os.environ.get("METAGPT_TEXT_TO_IMAGE_MODEL_URL") or model_url:
-        data = oas3_metagpt_text_to_image(text, size_type, model_url)
+        data = await oas3_metagpt_text_to_image(text, size_type, model_url)
         return image_declaration + data if data else ""
+
     if os.environ.get("OPENAI_API_KEY") or openai_api_key:
-        data = oas3_openai_text_to_image(text, size_type, openai_api_key)
+        data = await oas3_openai_text_to_image(text, size_type, openai_api_key)
         return image_declaration + data if data else ""
 
     raise EnvironmentError
diff --git a/metagpt/learn/text_to_speech.py b/metagpt/learn/text_to_speech.py
index 90dd878a1..e5eb3d488 100644
--- a/metagpt/learn/text_to_speech.py
+++ b/metagpt/learn/text_to_speech.py
@@ -16,7 +16,7 @@ from metagpt.utils.common import initialize_environment
 @skill_metadata(name="Text to speech",
                 description="Text-to-speech",
                 requisite="`AZURE_TTS_SUBSCRIPTION_KEY` and `AZURE_TTS_REGION`")
-def text_to_speech(text, lang="zh-CN", voice="zh-CN-XiaomoNeural", style="affectionate", role="Girl",
+async def text_to_speech(text, lang="zh-CN", voice="zh-CN-XiaomoNeural", style="affectionate", role="Girl",
                    subscription_key="", region="", **kwargs):
     """Text to speech
     For more details, check out:`https://learn.microsoft.com/en-us/azure/ai-services/speech-service/language-support?tabs=tts`
@@ -35,7 +35,7 @@ def text_to_speech(text, lang="zh-CN", voice="zh-CN-XiaomoNeural", style="affect
     audio_declaration = "data:audio/wav;base64,"
     if (os.environ.get("AZURE_TTS_SUBSCRIPTION_KEY") and os.environ.get("AZURE_TTS_REGION")) or \
             (subscription_key and region):
-        data = oas3_azsure_tts(text, lang, voice, style, role, subscription_key, region)
+        data = await oas3_azsure_tts(text, lang, voice, style, role, subscription_key, region)
         return audio_declaration + data if data else data
 
     raise EnvironmentError
diff --git a/metagpt/tools/azure_tts.py b/metagpt/tools/azure_tts.py
index 21e8f1b6c..1fd36e78c 100644
--- a/metagpt/tools/azure_tts.py
+++ b/metagpt/tools/azure_tts.py
@@ -6,6 +6,7 @@
 @File    : azure_tts.py
 @Desc    : azure TTS OAS3 api, which provides text-to-speech functionality
 """
+import asyncio
 from pathlib import Path
 from uuid import uuid4
 import base64
@@ -14,7 +15,7 @@ import sys
 sys.path.append(str(Path(__file__).resolve().parent.parent.parent))  # fix-bug: No module named 'metagpt'
 from metagpt.utils.common import initialize_environment
 from metagpt.logs import logger
-
+from aiofile import async_open
 from azure.cognitiveservices.speech import AudioConfig, SpeechConfig, SpeechSynthesizer
 import os
 
@@ -31,7 +32,7 @@ class AzureTTS:
         self.region = region if region else os.environ.get('AZURE_TTS_REGION')
 
     # 参数参考：https://learn.microsoft.com/zh-cn/azure/cognitive-services/speech-service/language-support?tabs=tts#voice-styles-and-roles
-    def synthesize_speech(self, lang, voice, text, output_file):
+    async def synthesize_speech(self, lang, voice, text, output_file):
         speech_config = SpeechConfig(
             subscription=self.subscription_key, region=self.region)
         speech_config.speech_synthesis_voice_name = voice
@@ -61,7 +62,7 @@ class AzureTTS:
 
 
 # Export
-def oas3_azsure_tts(text, lang="", voice="", style="", role="", subscription_key="", region=""):
+async def oas3_azsure_tts(text, lang="", voice="", style="", role="", subscription_key="", region=""):
     """Text to speech
     For more details, check out:`https://learn.microsoft.com/en-us/azure/ai-services/speech-service/language-support?tabs=tts`
 
@@ -95,9 +96,9 @@ def oas3_azsure_tts(text, lang="", voice="", style="", role="", subscription_key
     tts = AzureTTS(subscription_key=subscription_key, region=region)
     filename = Path(__file__).resolve().parent / (str(uuid4()).replace("-", "") + ".wav")
     try:
-        tts.synthesize_speech(lang=lang, voice=voice, text=xml_value, output_file=str(filename))
-        with open(str(filename), mode="rb") as reader:
-            data = reader.read()
+        await tts.synthesize_speech(lang=lang, voice=voice, text=xml_value, output_file=str(filename))
+        async with async_open(filename, mode="rb") as reader:
+            data = await reader.read()
             base64_string = base64.b64encode(data).decode('utf-8')
         filename.unlink()
     except Exception as e:
@@ -110,5 +111,7 @@ def oas3_azsure_tts(text, lang="", voice="", style="", role="", subscription_key
 if __name__ == "__main__":
     initialize_environment()
 
-    v = oas3_azsure_tts("测试，test")
+    loop = asyncio.new_event_loop()
+    v = loop.create_task(oas3_azsure_tts("测试，test"))
+    loop.run_until_complete(v)
     print(v)
diff --git a/metagpt/tools/hello.py b/metagpt/tools/hello.py
index e1bad6456..2eb4c31f0 100644
--- a/metagpt/tools/hello.py
+++ b/metagpt/tools/hello.py
@@ -17,7 +17,7 @@ import connexion
 
 
 # openapi implement
-def post_greeting(name: str) -> str:
+async def post_greeting(name: str) -> str:
     return f"Hello {name}\n"
 
 
diff --git a/metagpt/tools/metagpt_oas3_api_svc.py b/metagpt/tools/metagpt_oas3_api_svc.py
index 277d41dfb..624bb7d93 100644
--- a/metagpt/tools/metagpt_oas3_api_svc.py
+++ b/metagpt/tools/metagpt_oas3_api_svc.py
@@ -20,7 +20,7 @@ def oas_http_svc():
     """Start the OAS 3.0 OpenAPI HTTP service"""
     initialize_environment()
 
-    app = connexion.FlaskApp(__name__, specification_dir='../../.well-known/')
+    app = connexion.AioHttpApp(__name__, specification_dir='../../.well-known/')
     app.add_api("metagpt_oas3_api.yaml")
     app.add_api("openapi.yaml")
     app.run(port=8080)
diff --git a/metagpt/tools/metagpt_text_to_image.py b/metagpt/tools/metagpt_text_to_image.py
index 8588462d3..bc551134a 100644
--- a/metagpt/tools/metagpt_text_to_image.py
+++ b/metagpt/tools/metagpt_text_to_image.py
@@ -12,6 +12,7 @@ import sys
 from pathlib import Path
 from typing import List, Dict
 
+import aiohttp
 import requests
 from pydantic import BaseModel
 
@@ -27,7 +28,7 @@ class MetaGPTText2Image:
         """
         self.model_url = model_url if model_url else os.environ.get('METAGPT_TEXT_TO_IMAGE_MODEL')
 
-    def text_2_image(self, text, size_type="512x512"):
+    async def text_2_image(self, text, size_type="512x512"):
         """Text to image
 
         :param text: The text used for image conversion.
@@ -75,9 +76,9 @@ class MetaGPTText2Image:
             parameters: Dict
 
         try:
-            response = requests.post(self.model_url, headers=headers, json=data)
-            response.raise_for_status()  # Raise an exception for 4xx or 5xx responses
-            result = ImageResult(**response.json())
+            async with aiohttp.ClientSession() as session:
+                async with session.post(self.model_url, headers=headers, json=data) as response:
+                    result = ImageResult(**await response.json())
             if len(result.images) == 0:
                 return ""
             return result.images[0]
@@ -87,7 +88,7 @@ class MetaGPTText2Image:
 
 
 # Export
-def oas3_metagpt_text_to_image(text, size_type: str = "512x512", model_url=""):
+async def oas3_metagpt_text_to_image(text, size_type: str = "512x512", model_url=""):
     """Text to image
 
     :param text: The text used for image conversion.
@@ -99,7 +100,7 @@ def oas3_metagpt_text_to_image(text, size_type: str = "512x512", model_url=""):
         return ""
     if not model_url:
         model_url = os.environ.get('METAGPT_TEXT_TO_IMAGE_MODEL_URL')
-    return MetaGPTText2Image(model_url).text_2_image(text, size_type=size_type)
+    return await MetaGPTText2Image(model_url).text_2_image(text, size_type=size_type)
 
 
 if __name__ == "__main__":
diff --git a/metagpt/tools/openai_text_to_embedding.py b/metagpt/tools/openai_text_to_embedding.py
index 9eddd5bc1..119eb35b6 100644
--- a/metagpt/tools/openai_text_to_embedding.py
+++ b/metagpt/tools/openai_text_to_embedding.py
@@ -7,10 +7,12 @@
 @Desc    : OpenAI Text-to-Embedding OAS3 api, which provides text-to-embedding functionality.
             For more details, checkout: `https://platform.openai.com/docs/api-reference/embeddings/object`
 """
+import asyncio
 import os
 from pathlib import Path
 from typing import List
 
+import aiohttp
 import requests
 from pydantic import BaseModel
 import sys
@@ -47,7 +49,7 @@ class OpenAIText2Embedding:
         """
         self.openai_api_key = openai_api_key if openai_api_key else os.environ.get('OPENAI_API_KEY')
 
-    def text_2_embedding(self, text, model="text-embedding-ada-002"):
+    async def text_2_embedding(self, text, model="text-embedding-ada-002"):
         """Text to embedding
 
         :param text: The text used for embedding.
@@ -61,16 +63,16 @@ class OpenAIText2Embedding:
         }
         data = {"input": text, "model": model}
         try:
-            response = requests.post("https://api.openai.com/v1/embeddings", headers=headers, json=data)
-            response.raise_for_status()  # Raise an exception for 4xx or 5xx responses
-            return response.json()
+            async with aiohttp.ClientSession() as session:
+                async with session.post("https://api.openai.com/v1/embeddings", headers=headers, json=data) as response:
+                    return await response.json()
         except requests.exceptions.RequestException as e:
             logger.error(f"An error occurred:{e}")
         return {}
 
 
 # Export
-def oas3_openai_text_to_embedding(text, model="text-embedding-ada-002", openai_api_key=""):
+async def oas3_openai_text_to_embedding(text, model="text-embedding-ada-002", openai_api_key=""):
     """Text to embedding
 
     :param text: The text used for embedding.
@@ -82,11 +84,12 @@ def oas3_openai_text_to_embedding(text, model="text-embedding-ada-002", openai_a
         return ""
     if not openai_api_key:
         openai_api_key = os.environ.get("OPENAI_API_KEY")
-    return OpenAIText2Embedding(openai_api_key).text_2_embedding(text, model=model)
+    return await OpenAIText2Embedding(openai_api_key).text_2_embedding(text, model=model)
 
 
 if __name__ == "__main__":
     initialize_environment()
-
-    v = oas3_openai_text_to_embedding("Panda emoji")
+    loop = asyncio.new_event_loop()
+    v = loop.create_task(oas3_openai_text_to_embedding("Panda emoji"))
+    loop.run_until_complete(v)
     print(v)
diff --git a/metagpt/tools/openai_text_to_image.py b/metagpt/tools/openai_text_to_image.py
index 6ec96d166..cd48c62af 100644
--- a/metagpt/tools/openai_text_to_image.py
+++ b/metagpt/tools/openai_text_to_image.py
@@ -12,6 +12,7 @@ import sys
 from pathlib import Path
 from typing import List
 
+import aiohttp
 import requests
 from pydantic import BaseModel
 
@@ -27,7 +28,7 @@ class OpenAIText2Image:
         """
         self.openai_api_key = openai_api_key if openai_api_key else os.environ.get('OPENAI_API_KEY')
 
-    def text_2_image(self, text, size_type="1024x1024"):
+    async def text_2_image(self, text, size_type="1024x1024"):
         """Text to image
 
         :param text: The text used for image conversion.
@@ -48,27 +49,28 @@ class OpenAIText2Image:
         }
         data = {"prompt": text, "n": 1, "size": size_type}
         try:
-            response = requests.post("https://api.openai.com/v1/images/generations", headers=headers, json=data)
-            response.raise_for_status()  # Raise an exception for 4xx or 5xx responses
-            result = ImageResult(**response.json())
+            async with aiohttp.ClientSession() as session:
+                async with session.post("https://api.openai.com/v1/images/generations", headers=headers, json=data) as response:
+                    result = ImageResult(** await response.json())
         except requests.exceptions.RequestException as e:
             logger.error(f"An error occurred:{e}")
             return ""
         if len(result.data) > 0:
-            return OpenAIText2Image.get_image_data(result.data[0].url)
+            return await OpenAIText2Image.get_image_data(result.data[0].url)
         return ""
 
     @staticmethod
-    def get_image_data(url):
+    async def get_image_data(url):
         """Fetch image data from a URL and encode it as Base64
 
         :param url: Image url
         :return: Base64-encoded image data.
         """
         try:
-            response = requests.get(url)
-            response.raise_for_status()  # Raise an exception for 4xx or 5xx responses
-            image_data = response.content
+            async with aiohttp.ClientSession() as session:
+                async with session.get(url) as response:
+                    response.raise_for_status()  # 如果是 4xx 或 5xx 响应，会引发异常
+                    image_data = await response.read()
             base64_image = base64.b64encode(image_data).decode("utf-8")
             return base64_image
 
@@ -78,7 +80,7 @@ class OpenAIText2Image:
 
 
 # Export
-def oas3_openai_text_to_image(text, size_type: str = "1024x1024", openai_api_key=""):
+async def oas3_openai_text_to_image(text, size_type: str = "1024x1024", openai_api_key=""):
     """Text to image
 
     :param text: The text used for image conversion.
@@ -90,7 +92,7 @@ def oas3_openai_text_to_image(text, size_type: str = "1024x1024", openai_api_key
         return ""
     if not openai_api_key:
         openai_api_key = os.environ.get("OPENAI_API_KEY")
-    return OpenAIText2Image(openai_api_key).text_2_image(text, size_type=size_type)
+    return await OpenAIText2Image(openai_api_key).text_2_image(text, size_type=size_type)
 
 
 if __name__ == "__main__":
diff --git a/requirements.txt b/requirements.txt
index 70f2a3809..ed3f755c9 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -41,3 +41,4 @@ qdrant-client==1.4.0
 connexion[swagger-ui]
 aiohttp_jinja2
 azure-cognitiveservices-speech==1.30.0
+aiofile
\ No newline at end of file
diff --git a/tests/metagpt/learn/test_text_to_embedding.py b/tests/metagpt/learn/test_text_to_embedding.py
index c85e5dde8..d81a8ac1c 100644
--- a/tests/metagpt/learn/test_text_to_embedding.py
+++ b/tests/metagpt/learn/test_text_to_embedding.py
@@ -8,11 +8,11 @@
 """
 
 import asyncio
-import base64
 
 from pydantic import BaseModel
 
 from metagpt.learn.text_to_embedding import text_to_embedding
+from metagpt.tools.openai_text_to_embedding import ResultEmbedding
 
 
 async def mock_text_to_embedding():
@@ -25,7 +25,7 @@ async def mock_text_to_embedding():
 
     for i in inputs:
         seed = Input(**i)
-        data = text_to_embedding(seed.input)
+        data = await text_to_embedding(seed.input)
         v = ResultEmbedding(**data)
         assert len(v.data) > 0
 
diff --git a/tests/metagpt/learn/test_text_to_image.py b/tests/metagpt/learn/test_text_to_image.py
index 545c8a3ef..c359797de 100644
--- a/tests/metagpt/learn/test_text_to_image.py
+++ b/tests/metagpt/learn/test_text_to_image.py
@@ -25,10 +25,17 @@ async def mock_text_to_image():
 
     for i in inputs:
         seed = Input(**i)
-        base64_data = text_to_image(seed.input)
+        base64_data = await text_to_image(seed.input)
         assert base64_data != ""
         print(f"{seed.input} -> {base64_data}")
-        assert base64.b64decode(base64_data, validate=True)
+        flags = ";base64,"
+        assert flags in base64_data
+        ix = base64_data.find(flags) + len(flags)
+        declaration = base64_data[0: ix]
+        assert declaration
+        data = base64_data[ix:]
+        assert data
+        assert base64.b64decode(data, validate=True)
 
 
 def test_suite():
diff --git a/tests/metagpt/learn/test_text_to_speech.py b/tests/metagpt/learn/test_text_to_speech.py
index dbb599e38..68de5a3b2 100644
--- a/tests/metagpt/learn/test_text_to_speech.py
+++ b/tests/metagpt/learn/test_text_to_speech.py
@@ -24,10 +24,17 @@ async def mock_text_to_speech():
 
     for i in inputs:
         seed = Input(**i)
-        base64_data = text_to_speech(seed.input)
+        base64_data = await text_to_speech(seed.input)
         assert base64_data != ""
         print(f"{seed.input} -> {base64_data}")
-        assert base64.b64decode(base64_data, validate=True)
+        flags = ";base64,"
+        assert flags in base64_data
+        ix = base64_data.find(flags) + len(flags)
+        declaration = base64_data[0: ix]
+        assert declaration
+        data = base64_data[ix:]
+        assert data
+        assert base64.b64decode(data, validate=True)
 
 
 def test_suite():
diff --git a/tests/metagpt/tools/test_azure_tts.py b/tests/metagpt/tools/test_azure_tts.py
index 49dd7eed1..41d429109 100644
--- a/tests/metagpt/tools/test_azure_tts.py
+++ b/tests/metagpt/tools/test_azure_tts.py
@@ -7,6 +7,7 @@
 @Modified By: mashenquan, 2023-8-9, add more text formatting options
 @Modified By: mashenquan, 2023-8-17, move to `tools` folder.
 """
+import asyncio
 import sys
 from pathlib import Path
 
@@ -19,7 +20,7 @@ from metagpt.utils.common import initialize_environment
 def test_azure_tts():
     initialize_environment()
 
-    azure_tts = AzureTTS()
+    azure_tts = AzureTTS(subscription_key="", region="")
     text = """
         女儿看见父亲走了进来，问道：
             <mstts:express-as role="YoungAdultFemale" style="calm">
@@ -33,11 +34,13 @@ def test_azure_tts():
     path = WORKSPACE_ROOT / "tts"
     path.mkdir(exist_ok=True, parents=True)
     filename = path / "girl.wav"
-    result = azure_tts.synthesize_speech(
+    loop = asyncio.new_event_loop()
+    v = loop.create_task(azure_tts.synthesize_speech(
         lang="zh-CN",
         voice="zh-CN-XiaomoNeural",
         text=text,
-        output_file=str(filename))
+        output_file=str(filename)))
+    result = loop.run_until_complete(v)
 
     print(result)