Update return type

app.py

@@ -34,6 +34,8 @@ from preprocess.humanparsing.run_parsing import Parsing
 from preprocess.openpose.run_openpose import OpenPose
 from detectron2.data.detection_utils import convert_PIL_to_numpy,_apply_exif_orientation
 from torchvision.transforms.functional import to_pil_image
+import base64
+from io import BytesIO
 
 # For use ZeroGPU 
 # device = 'cuda' if torch.cuda.is_available() else 'cpu'
@@ -309,7 +311,13 @@ async def vton_run(
     target_cloth = Image.open(io.BytesIO(await upload_cloth.read()))
 
     results = quick_tryon(target_human, target_cloth, input_prompt)
-    return results[0]
+    
+    # Convert PIL Image to base64
+    buffered = BytesIO()
+    results[0].save(buffered, format="PNG")
+    img_str = base64.b64encode(buffered.getvalue()).decode()
+    
+    return {"image": img_str}
 
 
 # image_blocks = gr.Blocks().queue()