Update app.py

app.py

@@ -19,6 +19,14 @@ import torch
 from emu3.mllm.processing_emu3 import Emu3Processor
 import spaces
 
+import io
+import base64
+
+def image2str(image):
+    buf = io.BytesIO()
+    image.save(buf, format="WEBP")
+    i_str = base64.b64encode(buf.getvalue()).decode()
+    return f'<div style="float:left"><img src="data:image/png;base64, {i_str}"></div>'
 
 # Install flash attention, skipping CUDA build if necessary
 subprocess.run(
@@ -170,7 +178,7 @@ def chat(history, user_input, user_image):
         # Use Emu3-Chat for vision-language understanding
         response = vision_language_understanding(user_image, user_input)
         # Append the user input and response to the history
-        history = history + [(user_image, ""), (user_input, response)]
+        history = history + [(image2str(user_image) + user_input, response)]
     else:
         # Use Emu3-Gen for image generation
         generated_image = generate_image(user_input)