Spaces:
Runtime error
Runtime error
Download app.py from marcod9831/hallo2-final2: direct link, hf CLI and curl.
- Browser
- Download file 7.2 kB
-
https://huggingface.co/spaces/marcod9831/hallo2-final2/resolve/main/app.py
- Command line
-
hf download hf://spaces/marcod9831/hallo2-final2/app.py
-
curl -L -o app.py https://huggingface.co/spaces/marcod9831/hallo2-final2/resolve/main/app.py
7.2 kB
| """ | |
| Hallo2: Long-Duration and High-Resolution Audio-driven Portrait Image Animation | |
| Gradio App for HuggingFace Spaces | |
| """ | |
| import os | |
| import sys | |
| import argparse | |
| import tempfile | |
| import traceback | |
| from pathlib import Path | |
| import gradio as gr | |
| import torch | |
| # --------------------------------------------------------------------------- | |
| # Environment | |
| # --------------------------------------------------------------------------- | |
| ROOT_DIR = os.path.dirname(os.path.abspath(__file__)) | |
| sys.path.insert(0, ROOT_DIR) | |
| DEVICE = "cuda" if torch.cuda.is_available() else "cpu" | |
| print(f"Running on device: {DEVICE}") | |
| # --------------------------------------------------------------------------- | |
| # Model Download | |
| # --------------------------------------------------------------------------- | |
| def download_pretrained_models(): | |
| """ | |
| Download Hallo2 pretrained models if not already present. | |
| """ | |
| pretrained_dir = "./pretrained_models" | |
| os.makedirs(pretrained_dir, exist_ok=True) | |
| try: | |
| marker_file = os.path.join(pretrained_dir, ".download_complete") | |
| if os.path.exists(marker_file): | |
| print("Pretrained models already present.") | |
| return | |
| print("Downloading Hallo2 pretrained models...") | |
| from huggingface_hub import snapshot_download | |
| snapshot_download( | |
| repo_id="fudan-generative-ai/hallo2", | |
| local_dir=pretrained_dir, | |
| local_dir_use_symlinks=False, | |
| resume_download=True, | |
| ) | |
| with open(marker_file, "w") as f: | |
| f.write("ok") | |
| print("Model download completed.") | |
| except Exception as e: | |
| print(f"[WARNING] Model download failed:") | |
| print(str(e)) | |
| # --------------------------------------------------------------------------- | |
| # Inference | |
| # --------------------------------------------------------------------------- | |
| def run_inference( | |
| source_image_path, | |
| driving_audio_path, | |
| pose_weight, | |
| face_weight, | |
| lip_weight, | |
| face_expand_ratio, | |
| inference_steps, | |
| cfg_scale, | |
| ): | |
| try: | |
| from omegaconf import OmegaConf | |
| from scripts.inference_long import inference_process | |
| from hallo.utils.util import merge_videos | |
| output_dir = tempfile.mkdtemp(prefix="hallo2_out_") | |
| config_path = "configs/inference/long.yaml" | |
| if not os.path.exists(config_path): | |
| raise FileNotFoundError(f"Config not found: {config_path}") | |
| cfg = OmegaConf.load(config_path) | |
| cfg.inference_steps = int(inference_steps) | |
| cfg.cfg_scale = float(cfg_scale) | |
| cfg.save_path = output_dir | |
| cfg.source_image = source_image_path | |
| cfg.driving_audio = driving_audio_path | |
| cfg.pose_weight = float(pose_weight) | |
| cfg.face_weight = float(face_weight) | |
| cfg.lip_weight = float(lip_weight) | |
| cfg.face_expand_ratio = float(face_expand_ratio) | |
| temp_cfg = os.path.join(output_dir, "temp_config.yaml") | |
| OmegaConf.save(cfg, temp_cfg) | |
| args = argparse.Namespace( | |
| config=temp_cfg, | |
| source_image=source_image_path, | |
| driving_audio=driving_audio_path, | |
| pose_weight=float(pose_weight), | |
| face_weight=float(face_weight), | |
| lip_weight=float(lip_weight), | |
| face_expand_ratio=float(face_expand_ratio), | |
| audio_ckpt_dir=None, | |
| ) | |
| seg_path = inference_process(args) | |
| merged_path = os.path.join(Path(seg_path).parent, "merge_video.mp4") | |
| merge_videos(seg_path, merged_path) | |
| if os.path.exists(merged_path): | |
| return merged_path | |
| mp4_files = list(Path(seg_path).glob("*.mp4")) | |
| if mp4_files: | |
| return str(mp4_files[0]) | |
| raise RuntimeError("No output video generated.") | |
| except Exception: | |
| raise RuntimeError(traceback.format_exc()) | |
| # --------------------------------------------------------------------------- | |
| # Gradio Inference Wrapper | |
| # --------------------------------------------------------------------------- | |
| def infer( | |
| source_image, | |
| driving_audio, | |
| pose_weight, | |
| face_weight, | |
| lip_weight, | |
| face_expand_ratio, | |
| inference_steps, | |
| cfg_scale, | |
| ): | |
| if source_image is None: | |
| return None, "Please upload a portrait image." | |
| if driving_audio is None: | |
| return None, "Please upload a WAV audio file." | |
| try: | |
| video_path = run_inference( | |
| source_image_path=source_image, | |
| driving_audio_path=driving_audio, | |
| pose_weight=pose_weight, | |
| face_weight=face_weight, | |
| lip_weight=lip_weight, | |
| face_expand_ratio=face_expand_ratio, | |
| inference_steps=inference_steps, | |
| cfg_scale=cfg_scale, | |
| ) | |
| return video_path, "Video generated successfully." | |
| except Exception as e: | |
| return None, f"Error:\n{str(e)}" | |
| # --------------------------------------------------------------------------- | |
| # UI | |
| # --------------------------------------------------------------------------- | |
| with gr.Blocks(title="Hallo2 Audio Driven Portrait Animation") as demo: | |
| gr.Markdown( | |
| """ | |
| # Hallo2 | |
| Upload: | |
| - Portrait image | |
| - WAV audio | |
| Generate realistic talking-head animation. | |
| """ | |
| ) | |
| with gr.Row(): | |
| with gr.Column(): | |
| source_image = gr.Image(label="Portrait Image", type="filepath") | |
| driving_audio = gr.Audio(label="Driving Audio", type="filepath") | |
| with gr.Accordion("Advanced Settings", open=False): | |
| pose_weight = gr.Slider( | |
| 0.0, 2.0, value=1.0, step=0.1, label="Pose Weight" | |
| ) | |
| face_weight = gr.Slider( | |
| 0.0, 2.0, value=1.0, step=0.1, label="Face Weight" | |
| ) | |
| lip_weight = gr.Slider( | |
| 0.0, 2.0, value=1.0, step=0.1, label="Lip Weight" | |
| ) | |
| face_expand_ratio = gr.Slider( | |
| 1.0, 2.0, value=1.2, step=0.05, label="Face Expand Ratio" | |
| ) | |
| inference_steps = gr.Slider( | |
| 10, 100, value=40, step=5, label="Inference Steps" | |
| ) | |
| cfg_scale = gr.Slider( | |
| 1.0, 10.0, value=3.5, step=0.5, label="CFG Scale" | |
| ) | |
| run_btn = gr.Button("Generate Video", variant="primary") | |
| with gr.Column(): | |
| output_video = gr.Video(label="Generated Video") | |
| status_box = gr.Textbox(label="Status", lines=8, interactive=False) | |
| run_btn.click( | |
| fn=infer, | |
| inputs=[ | |
| source_image, | |
| driving_audio, | |
| pose_weight, | |
| face_weight, | |
| lip_weight, | |
| face_expand_ratio, | |
| inference_steps, | |
| cfg_scale, | |
| ], | |
| outputs=[output_video, status_box], | |
| ) | |
| # --------------------------------------------------------------------------- | |
| # Startup | |
| # --------------------------------------------------------------------------- | |
| try: | |
| download_pretrained_models() | |
| except Exception as e: | |
| print(f"Startup warning: {e}") | |
| if __name__ == "__main__": | |
| demo.queue(max_size=5).launch() | |