zeroscope-v2 / app.py
hysts's picture
hysts HF staff
Update
38666fe
#!/usr/bin/env python
from __future__ import annotations
import os
import random
import tempfile
import gradio as gr
import imageio
import numpy as np
import spaces
import torch
from diffusers import DiffusionPipeline, DPMSolverMultistepScheduler
DESCRIPTION = "# zeroscope v2"
if not torch.cuda.is_available():
DESCRIPTION += "\n<p>Running on CPU 🥶 This demo does not work on CPU.</p>"
MAX_NUM_FRAMES = int(os.getenv("MAX_NUM_FRAMES", "200"))
DEFAULT_NUM_FRAMES = min(MAX_NUM_FRAMES, int(os.getenv("DEFAULT_NUM_FRAMES", "24")))
MAX_SEED = np.iinfo(np.int32).max
if torch.cuda.is_available():
pipe = DiffusionPipeline.from_pretrained("cerspense/zeroscope_v2_576w", torch_dtype=torch.float16)
pipe.scheduler = DPMSolverMultistepScheduler.from_config(pipe.scheduler.config)
pipe.enable_model_cpu_offload()
pipe.enable_vae_slicing()
def randomize_seed_fn(seed: int, randomize_seed: bool) -> int:
if randomize_seed:
seed = random.randint(0, MAX_SEED)
return seed
def to_video(frames: np.ndarray, fps: int) -> str:
frames = np.clip((frames * 255), 0, 255).astype(np.uint8)
out_file = tempfile.NamedTemporaryFile(suffix=".mp4", delete=False)
writer = imageio.get_writer(out_file.name, format="FFMPEG", fps=fps)
for frame in frames:
writer.append_data(frame)
writer.close()
return out_file.name
@spaces.GPU
def generate(
prompt: str,
seed: int,
num_frames: int,
num_inference_steps: int,
progress=gr.Progress(track_tqdm=True),
) -> str:
generator = torch.Generator().manual_seed(seed)
frames = pipe(
prompt,
num_inference_steps=num_inference_steps,
num_frames=num_frames,
width=576,
height=320,
generator=generator,
).frames[0]
return to_video(frames, 8)
examples = [
["An astronaut riding a horse", 0, 24, 25],
["A panda eating bamboo on a rock", 0, 24, 25],
["Spiderman is surfing", 0, 24, 25],
]
with gr.Blocks(css="style.css") as demo:
gr.Markdown(DESCRIPTION)
gr.DuplicateButton(
value="Duplicate Space for private use",
elem_id="duplicate-button",
visible=os.getenv("SHOW_DUPLICATE_BUTTON") == "1",
)
with gr.Group():
with gr.Row():
prompt = gr.Text(
label="Prompt",
show_label=False,
max_lines=1,
placeholder="Enter your prompt",
container=False,
)
run_button = gr.Button("Generate video", scale=0)
result = gr.Video(label="Result", show_label=False)
with gr.Accordion("Advanced options", open=False):
seed = gr.Slider(
label="Seed",
minimum=0,
maximum=MAX_SEED,
step=1,
value=0,
)
randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
num_frames = gr.Slider(
label="Number of frames",
minimum=24,
maximum=MAX_NUM_FRAMES,
step=1,
value=24,
info="Note that the content of the video also changes when you change the number of frames.",
)
num_inference_steps = gr.Slider(
label="Number of inference steps",
minimum=10,
maximum=50,
step=1,
value=25,
)
inputs = [
prompt,
seed,
num_frames,
num_inference_steps,
]
gr.Examples(
examples=examples,
inputs=inputs,
outputs=result,
fn=generate,
)
gr.on(
triggers=[prompt.submit, run_button.click],
fn=randomize_seed_fn,
inputs=[seed, randomize_seed],
outputs=seed,
queue=False,
api_name=False,
).then(
fn=generate,
inputs=inputs,
outputs=result,
api_name="run",
)
if __name__ == "__main__":
demo.queue(max_size=10).launch()