File size: 5,216 Bytes
d24fe1c 017c8fa |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 |
import os, sys
import gradio as gr
import torch
import shutil
from src.gradio_demo import SadTalker
def sadtalker_demo():
# For Hugging Face, we'll use the current directory structure
checkpoint_path = 'checkpoints'
config_path = 'src/config'
try:
sad_talker = SadTalker(checkpoint_path, config_path, lazy_load=True)
except Exception as e:
print(f"Warning: Could not initialize SadTalker: {e}")
sad_talker = None
def generate_video(source_image, driven_audio, preprocess_type, is_still_mode, enhancer, batch_size, size_of_image, pose_style):
if sad_talker is None:
return "Error: SadTalker not initialized. Please ensure all model files are uploaded."
try:
return sad_talker.test(
source_image=source_image,
driven_audio=driven_audio,
preprocess=preprocess_type,
still_mode=is_still_mode,
use_enhancer=enhancer,
batch_size=batch_size,
size=size_of_image,
pose_style=pose_style
)
except Exception as e:
return f"Error generating video: {str(e)}"
with gr.Row().style(equal_height=False):
with gr.Column(variant='panel'):
with gr.Tabs(elem_id="sadtalker_source_image"):
with gr.TabItem('Upload image'):
with gr.Row():
source_image = gr.Image(
label="Source image",
source="upload",
type="filepath",
elem_id="img2img_image"
).style(width=512)
with gr.Tabs(elem_id="sadtalker_driven_audio"):
with gr.TabItem('Upload Audio'):
with gr.Column(variant='panel'):
driven_audio = gr.Audio(
label="Input audio",
source="upload",
type="filepath"
)
with gr.Column(variant='panel'):
with gr.Tabs(elem_id="sadtalker_checkbox"):
with gr.TabItem('Settings'):
gr.Markdown("""
Need help? Please visit our [best practice page](https://github.com/OpenTalker/SadTalker/blob/main/docs/best_practice.md) for more details
""")
with gr.Column(variant='panel'):
pose_style = gr.Slider(
minimum=0,
maximum=46,
step=1,
label="Pose style",
value=0
)
size_of_image = gr.Radio(
[256, 512],
value=256,
label='Face model resolution',
info="Use 256/512 model?"
)
preprocess_type = gr.Radio(
['crop', 'resize','full', 'extcrop', 'extfull'],
value='crop',
label='preprocess',
info="How to handle input image?"
)
is_still_mode = gr.Checkbox(
label="Still Mode (fewer head motion, works with preprocess `full`)"
)
batch_size = gr.Slider(
label="Batch size in generation",
step=1,
maximum=10,
value=2
)
enhancer = gr.Checkbox(
label="GFPGAN as Face enhancer"
)
submit = gr.Button(
'Generate',
elem_id="sadtalker_generate",
variant='primary'
)
with gr.Tabs(elem_id="sadtalker_generated"):
gen_video = gr.Video(
label="Generated video",
format="mp4"
).style(width=512)
submit.click(
fn=generate_video,
inputs=[
source_image,
driven_audio,
preprocess_type,
is_still_mode,
enhancer,
batch_size,
size_of_image,
pose_style
],
outputs=[gen_video]
)
return sadtalker_interface
if __name__ == "__main__":
demo = sadtalker_demo()
demo.queue()
demo.launch() |