vidcraft / app_local.py
ihsanvp's picture
fix: local with spaces
448a859
raw
history blame
2.35 kB
import gradio as gr
import torch
import torchvision
from diffusers import I2VGenXLPipeline, DiffusionPipeline
from torchvision.transforms.functional import to_tensor
from PIL import Image
if gr.NO_RELOAD:
n_steps = 40
high_noise_frac = 0.8
negative_prompt = "Distorted, discontinuous, Ugly, blurry, low resolution, motionless, static, disfigured, disconnected limbs, Ugly faces, incomplete arms"
generator = torch.manual_seed(8888)
base = DiffusionPipeline.from_pretrained(
"stabilityai/stable-diffusion-xl-base-1.0",
torch_dtype=torch.float16,
variant="fp16",
use_safetensors=True,
)
# refiner = DiffusionPipeline.from_pretrained(
# "stabilityai/stable-diffusion-xl-refiner-1.0",
# text_encoder_2=base.text_encoder_2,
# vae=base.vae,
# torch_dtype=torch.float16,
# use_safetensors=True,
# variant="fp16",
# )
# refiner.to("cuda")
# base.to("cuda")
# refiner.enable_model_cpu_offload()
base.enable_model_cpu_offload()
pipeline = I2VGenXLPipeline.from_pretrained("ali-vilab/i2vgen-xl", torch_dtype=torch.float16, variant="fp16")
pipeline.enable_model_cpu_offload()
pipeline.unet.enable_forward_chunking()
def generate(prompt: str):
image = base(
prompt=prompt,
num_inference_steps=n_steps,
# denoising_end=high_noise_frac,
# output_type="latent",
).images[0]
# image = refiner(
# prompt=prompt,
# num_inference_steps=n_steps,
# denoising_start=high_noise_frac,
# image=image,
# ).images[0]
# print(image)
# print(type(image))
# print(image.size())
image.save("frame.jpg")
image = to_tensor(image)
frames: list[Image.Image] = pipeline(
prompt=prompt,
image=image,
num_inference_steps=50,
negative_prompt=negative_prompt,
guidance_scale=9.0,
generator=generator,
decode_chunk_size=6,
).frames[0]
frames = [to_tensor(frame.convert("RGB")).mul(255).byte().permute(1, 2, 0) for frame in frames]
frames = torch.stack(frames)
torchvision.io.write_video("video.mp4", frames, fps=4)
return "video.mp4"
app = gr.Interface(
fn=generate,
inputs=["text"],
outputs=gr.Video()
)
if __name__ == "__main__":
app.launch()