diff --git a/lama_cleaner/app/src/components/Editor/Editor.tsx b/lama_cleaner/app/src/components/Editor/Editor.tsx
index 5d0b0ee..2c2a697 100644
--- a/lama_cleaner/app/src/components/Editor/Editor.tsx
+++ b/lama_cleaner/app/src/components/Editor/Editor.tsx
@@ -598,6 +598,15 @@ export default function Editor() {
     }
   }, [runRenderablePlugin])
 
+  useEffect(() => {
+    emitter.on(PluginName.GFPGAN, () => {
+      runRenderablePlugin(PluginName.GFPGAN)
+    })
+    return () => {
+      emitter.off(PluginName.GFPGAN)
+    }
+  }, [runRenderablePlugin])
+
   useEffect(() => {
     emitter.on(PluginName.RealESRGAN, (data: any) => {
       runRenderablePlugin(PluginName.RealESRGAN, data)
diff --git a/lama_cleaner/app/src/components/Plugins/Plugins.tsx b/lama_cleaner/app/src/components/Plugins/Plugins.tsx
index 289ae19..087c4ac 100644
--- a/lama_cleaner/app/src/components/Plugins/Plugins.tsx
+++ b/lama_cleaner/app/src/components/Plugins/Plugins.tsx
@@ -4,6 +4,7 @@ import { CursorArrowRaysIcon, GifIcon } from '@heroicons/react/24/outline'
 import {
   BoxModelIcon,
   ChevronRightIcon,
+  FaceIcon,
   HobbyKnifeIcon,
   MixIcon,
 } from '@radix-ui/react-icons'
@@ -20,6 +21,7 @@ import Button from '../shared/Button'
 export enum PluginName {
   RemoveBG = 'RemoveBG',
   RealESRGAN = 'RealESRGAN',
+  GFPGAN = 'GFPGAN',
   InteractiveSeg = 'InteractiveSeg',
   MakeGIF = 'MakeGIF',
 }
@@ -33,6 +35,10 @@ const pluginMap = {
     IconClass: BoxModelIcon,
     showName: 'RealESRGAN 4x',
   },
+  [PluginName.GFPGAN]: {
+    IconClass: FaceIcon,
+    showName: 'GFPGAN',
+  },
   [PluginName.InteractiveSeg]: {
     IconClass: CursorArrowRaysIcon,
     showName: 'Interactive Seg',
diff --git a/lama_cleaner/server.py b/lama_cleaner/server.py
index 99b5958..cb79b8e 100644
--- a/lama_cleaner/server.py
+++ b/lama_cleaner/server.py
@@ -327,28 +327,46 @@ def run_plugin():
         return "Plugin not found", 500
 
     origin_image_bytes = files["image"].read()  # RGB
-    rgb_np_img, _ = load_img(origin_image_bytes)
+    rgb_np_img, alpha_channel, exif = load_img(origin_image_bytes, return_exif=True)
 
     start = time.time()
-    res = plugins[name](rgb_np_img, files, form)
+    bgr_res = plugins[name](rgb_np_img, files, form)
     logger.info(f"{name} process time: {(time.time() - start) * 1000}ms")
     torch_gc()
 
     if name == MakeGIF.name:
-        filename = form["filename"]
         return send_file(
-            io.BytesIO(res),
+            io.BytesIO(bgr_res),
             mimetype="image/gif",
             as_attachment=True,
-            attachment_filename=filename,
+            attachment_filename=form["filename"],
         )
+
+    if name == RemoveBG.name:
+        rgb_res = cv2.cvtColor(bgr_res, cv2.COLOR_BGRA2RGBA)
+        ext = "png"
     else:
-        response = make_response(
-            send_file(
-                io.BytesIO(numpy_to_bytes(res, "png")),
-                mimetype=f"image/png",
+        rgb_res = cv2.cvtColor(bgr_res, cv2.COLOR_BGR2RGB)
+        ext = get_image_ext(origin_image_bytes)
+        if alpha_channel is not None:
+            if alpha_channel.shape[:2] != rgb_res.shape[:2]:
+                alpha_channel = cv2.resize(
+                    alpha_channel, dsize=(rgb_res.shape[1], rgb_res.shape[0])
+                )
+            rgb_res = np.concatenate(
+                (rgb_res, alpha_channel[:, :, np.newaxis]), axis=-1
             )
+
+    response = make_response(
+        send_file(
+            io.BytesIO(
+                pil_to_bytes(
+                    Image.fromarray(rgb_res), ext, quality=image_quality, exif=exif
+                )
+            ),
+            mimetype=f"image/{ext}",
         )
+    )
     return response
 
 
diff --git a/lama_cleaner/tests/test_model.py b/lama_cleaner/tests/test_model.py
index d1af506..a76f3c6 100644
--- a/lama_cleaner/tests/test_model.py
+++ b/lama_cleaner/tests/test_model.py
@@ -1,4 +1,3 @@
-import os
 from pathlib import Path
 
 import cv2
diff --git a/lama_cleaner/tests/test_plugins.py b/lama_cleaner/tests/test_plugins.py
index 0634a5d..6d68ec9 100644
--- a/lama_cleaner/tests/test_plugins.py
+++ b/lama_cleaner/tests/test_plugins.py
@@ -1,27 +1,46 @@
 from pathlib import Path
 
 import cv2
+import pytest
+import torch.cuda
 
-from lama_cleaner.plugins import RemoveBG, RealESRGANUpscaler
+from lama_cleaner.plugins import RemoveBG, RealESRGANUpscaler, GFPGANPlugin
 
 current_dir = Path(__file__).parent.absolute().resolve()
 save_dir = current_dir / "result"
 save_dir.mkdir(exist_ok=True, parents=True)
 img_p = current_dir / "bunny.jpeg"
+bgr_img = cv2.imread(str(img_p))
+rgb_img = cv2.cvtColor(bgr_img, cv2.COLOR_BGR2RGB)
+
+
+def _save(img, name):
+    cv2.imwrite(str(save_dir / name), img)
 
 
 def test_remove_bg():
     model = RemoveBG()
-    img = cv2.imread(str(img_p))
-    res = model.forward(img)
-    cv2.imwrite(str(save_dir / "test_remove_bg.png"), res)
+    res = model.forward(bgr_img)
+    _save(res, "test_remove_bg.png")
 
 
-def test_upscale():
-    model = RealESRGANUpscaler("cpu")
-    img = cv2.imread(str(img_p))
-    res = model.forward(img, 2)
-    cv2.imwrite(str(save_dir / "test_upscale_x2.png"), res)
+@pytest.mark.parametrize("device", ["cuda", "cpu"])
+def test_upscale(device):
+    if device == "cuda" and not torch.cuda.is_available():
+        return
 
-    res = model.forward(img, 4)
-    cv2.imwrite(str(save_dir / "test_upscale_x4.png"), res)
+    model = RealESRGANUpscaler("realesr-general-x4v3", device)
+    res = model.forward(bgr_img, 2)
+    _save(res, "test_upscale_x2.png")
+
+    res = model.forward(bgr_img, 4)
+    _save(res, "test_upscale_x4.png")
+
+
+@pytest.mark.parametrize("device", ["cuda", "cpu"])
+def test_gfpgan(device):
+    if device == "cuda" and not torch.cuda.is_available():
+        return
+    model = GFPGANPlugin(device)
+    res = model(rgb_img, None, None)
+    _save(res, "test_gfpgan.png")