diff --git a/.gitattributes b/.gitattributes
index a6344aac8c09253b3b630fb776ae94478aa0275b..7307b16ff423aa171ac31571fa774436a3ed6b9f 100644
--- a/.gitattributes
+++ b/.gitattributes
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
*.zip filter=lfs diff=lfs merge=lfs -text
*.zst filter=lfs diff=lfs merge=lfs -text
*tfevents* filter=lfs diff=lfs merge=lfs -text
+difpoint/src/utils/dependencies/insightface/data/images/t1.jpg filter=lfs diff=lfs merge=lfs -text
diff --git a/app.py b/app.py
new file mode 100644
index 0000000000000000000000000000000000000000..b6df7955417669042688751f1acf6afa956d3d82
--- /dev/null
+++ b/app.py
@@ -0,0 +1,124 @@
+import os, sys
+import gradio as gr
+from difpoint.inference import Inferencer
+from TTS.api import TTS
+import torch
+import time
+from flask import send_from_directory
+from huggingface_hub import snapshot_download
+import spaces
+import tensorrt
+import multiprocessing as mp
+import pickle
+mp.set_start_method('spawn', force=True)
+
+
+repo_id = "ChaolongYang/KDTalker"
+local_dir = "./downloaded_repo"
+snapshot_download(repo_id=repo_id, local_dir=local_dir)
+print("\nFiles downloaded:")
+for root, dirs, files in os.walk(local_dir):
+ for file in files:
+ file_path = os.path.join(root, file)
+ print(file_path)
+
+result_dir = "results"
+def set_upload():
+ return "upload"
+def set_microphone():
+ return "microphone"
+def set_tts():
+ return "tts"
+def create_kd_talker():
+ return Inferencer()
+
+@spaces.GPU
+def predict(prompt, upload_reference_audio, microphone_reference_audio, reference_audio_type):
+ global result_dir
+ output_file_path = os.path.join('./downloaded_repo/', 'output.wav')
+ if reference_audio_type == 'upload':
+ audio_file_pth = upload_reference_audio
+ elif reference_audio_type == 'microphone':
+ audio_file_pth = microphone_reference_audio
+ tts = TTS('tts_models/multilingual/multi-dataset/your_tts')
+ tts.tts_to_file(
+ text=prompt,
+ file_path=output_file_path,
+ speaker_wav=audio_file_pth,
+ language="en",
+ )
+ return gr.Audio(value=output_file_path, type='filepath')
+
+@spaces.GPU
+def generate(upload_driven_audio, tts_driven_audio, driven_audio_type, source_image, smoothed_pitch, smoothed_yaw, smoothed_roll, smoothed_t):
+ return Inferencer().generate_with_audio_img(upload_driven_audio, tts_driven_audio, driven_audio_type, source_image,
+ smoothed_pitch, smoothed_yaw, smoothed_roll, smoothed_t)
+
+
+def main():
+ if torch.cuda.is_available():
+ device = "cuda"
+ else:
+ device = "cpu"
+
+ with gr.Blocks(analytics_enabled=False) as interface:
+ gr.Markdown(
+ """
+
+
Unlock Pose Diversity: Accurate and Efficient Implicit Keypoint-based Spatiotemporal Diffusion for Audio-driven Talking Portrait
+
+
+
+
+