diff --git a/data/avatars/wav2lip_avatar2/face_imgs/00000000.png b/data/avatars/wav2lip_avatar2/face_imgs/00000000.png
index 9bc012e..2d31bb9 100644
Binary files a/data/avatars/wav2lip_avatar2/face_imgs/00000000.png and b/data/avatars/wav2lip_avatar2/face_imgs/00000000.png differ
diff --git a/human/audio_inference_handler.py b/human/audio_inference_handler.py
index 30e430b..6c78c84 100644
--- a/human/audio_inference_handler.py
+++ b/human/audio_inference_handler.py
@@ -59,7 +59,7 @@ class AudioInferenceHandler(AudioHandler):
         super().on_message(message)
 
     def __on_run(self):
-        wav2lip_path = os.path.join(current_file_path, '..', 'checkpoints', 'wav2lip_gan.pth')
+        wav2lip_path = os.path.join(current_file_path, '..', 'checkpoints', 'wav2lip.pth')
         logger.info(f'AudioInferenceHandler init, path:{wav2lip_path}')
         model = load_model(wav2lip_path)
         logger.info("Model loaded")
@@ -130,7 +130,7 @@ class AudioInferenceHandler(AudioHandler):
 
                     img_batch = torch.FloatTensor(np.transpose(img_batch, (0, 3, 1, 2))).to(device)
                     mel_batch = torch.FloatTensor(np.transpose(mel_batch, (0, 3, 1, 2))).to(device)
-                    # print('img_batch:', img_batch.shape, 'mel_batch:', mel_batch.shape)
+                    print('img_batch:', img_batch.shape, 'mel_batch:', mel_batch.shape)
                     with torch.no_grad():
                         pred = model(mel_batch, img_batch)
 
diff --git a/human/human_context.py b/human/human_context.py
index b0c5d37..40eee13 100644
--- a/human/human_context.py
+++ b/human/human_context.py
@@ -18,7 +18,7 @@ current_file_path = os.path.dirname(os.path.abspath(__file__))
 class HumanContext:
     def __init__(self):
         self._fps = 50  # 20 ms per frame
-        self._image_size = 128
+        self._image_size = 288
         self._batch_size = 16
         self._sample_rate = 16000
         self._stride_left_size = 10