add video

app.py

@@ -40,7 +40,8 @@ from langchain.vectorstores.faiss import FAISS
 from langchain.docstore.document import Document
 from langchain.chains.question_answering import load_qa_chain
 import azure.cognitiveservices.speech as speechsdk
- 
+import base64
+
 
 news_api_key = os.environ["NEWS_API_KEY"]
 
@@ -56,7 +57,7 @@ MAX_TOKENS = 512
 
 LOOPING_TALKING_HEAD = "videos/Michelle.mp4"
 TALKING_HEAD_WIDTH = "192"
-MAX_TALKING_HEAD_TEXT_LENGTH = 155
+MAX_TALKING_HEAD_TEXT_LENGTH = 20
 
 # Pertains to Express-inator functionality
 NUM_WORDS_DEFAULT = 0
@@ -431,7 +432,9 @@ class ChatWrapper:
             if speak_text:
                 if talking_head:
                     if len(output) <= MAX_TALKING_HEAD_TEXT_LENGTH:
-                        html_video, temp_file = do_html_video_speak(output, translate_to)
+                        # html_video, temp_file = do_html_video_speak(output, translate_to)
+                        html_audio, temp_aud_file = do_html_audio_speak_azure(output, translate_to)
+                        html_video, temp_file = do_html_video_speak_sad_talker(output, translate_to)
                     else:
                         temp_file = LOOPING_TALKING_HEAD
                         html_video = create_html_video(temp_file, TALKING_HEAD_WIDTH)
@@ -552,6 +555,48 @@ def create_html_video(file_name, width):
     html_video = f'<video width={width} height={width} autoplay muted loop><source src={temp_file_url} type="video/mp4" poster="Michelle.png"></video>'
     return html_video
 
+def ToBase64(file):
+    with open(file, 'rb') as fileObj:
+        image_data = fileObj.read()
+        base64_data = base64.b64encode(image_data)s
+        return base64_data.decode()
+
+
+def do_html_video_speak_sad_talker(temp_aud_file, azure_language):
+    
+    GRADIO_URL=os.environ["GRADIO_URL"],
+
+    img_data = ToBase64("images/Masahiro.png")
+    audio_data = ToBase64(temp_aud_file)
+
+    response = requests.post(GRADIO_URL+"/run/sad_talker", json={
+        "data": [
+            "data:image/png;base64,"+img_data,
+            {"name":"audio.wav","data":"data:audio/wav;base64,"+audio_data},
+            "crop",
+            False,
+            False,
+        ]
+    },timeout=3000)
+    print(response.text)
+    res = response.json()
+
+    data = res["data"]
+    print(data)
+    video_rul = GRADIO_URL+"/file=" + data[0][0]['name']
+    print(video_rul)
+    
+    html_video = '<pre>no video</pre>'
+
+    # with open('videos/tempfile.mp4', 'wb') as f:
+    #     f.write(response_stream.read())
+    # temp_file = gr.File("videos/tempfile.mp4")
+    # temp_file_url = "/file=" + temp_file.value['name']
+    temp_file_url=video_rul
+    html_video = f'<video width={TALKING_HEAD_WIDTH} height={TALKING_HEAD_WIDTH} autoplay><source src={temp_file_url} type="video/mp4" poster="Michelle.png"></video>'
+
+    return html_video, "videos/tempfile.mp4"
+
 
 def do_html_video_speak(words_to_speak, azure_language):
     azure_voice = AZURE_VOICE_DATA.get_voice(azure_language, "Male")