Test / app.py
pranavajay's picture
Update app.py
fba1156 verified
import gradio as gr
import torch
import os
import spaces
import uuid
from diffusers import AnimateDiffPipeline, MotionAdapter, EulerDiscreteScheduler
from diffusers.utils import export_to_video
from huggingface_hub import hf_hub_download
from safetensors.torch import load_file
from PIL import Image
# Constants
bases = {
"ToonYou": "frankjoshua/toonyou_beta6",
"epiCRealism": "emilianJR/epiCRealism"
}
step_loaded = None
base_loaded = "ToonYou"
motion_loaded = None
# Ensure model and scheduler are initialized in GPU-enabled function
if not torch.cuda.is_available():
raise NotImplementedError("No GPU detected!")
device = "cuda"
dtype = torch.float16
pipe = AnimateDiffPipeline.from_pretrained(bases[base_loaded], torch_dtype=dtype).to(device)
pipe.scheduler = EulerDiscreteScheduler.from_config(pipe.scheduler.config, timestep_spacing="trailing", beta_schedule="linear")
# Safety checkers
from transformers import CLIPFeatureExtractor
feature_extractor = CLIPFeatureExtractor.from_pretrained("openai/clip-vit-base-patch32")
# Function
@spaces.GPU(enable_queue=True)
def generate_image(prompt, base, motion, step, progress=gr.Progress()):
global step_loaded
global base_loaded
global motion_loaded
print(prompt, base, step)
if base_loaded != base:
pipe.unet.load_state_dict(torch.load(hf_hub_download(bases[base], "unet/diffusion_pytorch_model.bin"), map_location=device), strict=False)
base_loaded = base
if motion_loaded != motion:
pipe.unload_lora_weights()
if motion != "":
pipe.load_lora_weights(motion, adapter_name="motion")
pipe.set_adapters(["motion"], [0.7])
motion_loaded = motion
output = pipe(prompt=prompt, guidance_scale=1.0, num_inference_steps=int(step))
name = str(uuid.uuid4()).replace("-", "")
path = f"/tmp/{name}.mp4"
export_to_video(output.frames[0], path, fps=10)
return path
# Gradio Interface
with gr.Blocks(css="style.css") as demo:
gr.HTML(
"<h1><center>Text To Video ⚡</center></h1>" +
"<p><center>SDXL text-to-video generation</center></p>" +
"<p><center><a href='https://github/ByteDance/AnimateDiff-Lightning'>https://huggingface.co/ByteDance/AnimateDiff-Lightning</a></center></p>"
)
with gr.Group():
with gr.Row():
prompt = gr.Textbox(
label='Prompt (Eng)'
)
with gr.Row():
select_base = gr.Dropdown(
label='Base model',
choices=[
"ToonYou",
"epiCRealism",
],
value=base_loaded,
interactive=True
)
select_motion = gr.Dropdown(
label='Motion',
choices=[
("Default", ""),
("Zoom in", "guoyww/animatediff-motion-lora-zoom-in"),
("Zoom out", "guoyww/animatediff-motion-lora-zoom-out"),
("Tilt up", "guoyww/animatediff-motion-lora-tilt-up"),
("Tilt down", "guoyww/animatediff-motion-lora-tilt-down"),
("Pan left", "guoyww/animatediff-motion-lora-pan-left"),
("Pan right", "guoyww/animatediff-motion-lora-pan-right"),
("Roll left", "guoyww/animatediff-motion-lora-rolling-anticlockwise"),
("Roll right", "guoyww/animatediff-motion-lora-rolling-clockwise"),
],
value="",
interactive=True
)
select_step = gr.Dropdown(
label='Inference steps',
choices=[
('1-Step', 1),
('2-Step', 2),
('4-Step', 4),
('8-Step', 8),
('16-Step', 16),
('25-Step', 25)],
value=16,
interactive=True
)
submit = gr.Button(
scale=1,
variant='primary'
)
video = gr.Video(
label='Text To Video',
autoplay=True,
height=512,
width=512,
elem_id="video_output"
)
prompt.submit(
fn=generate_image,
inputs=[prompt, select_base, select_motion, select_step],
outputs=video,
)
submit.click(
fn=generate_image,
inputs=[prompt, select_base, select_motion, select_step],
outputs=video,
)
demo.queue().launch()