|
import os |
|
import gradio as gr |
|
from loguru import logger |
|
|
|
|
|
def download_models(): |
|
logger.info("Scaricamento dei modelli...") |
|
os.system("apt update && apt install aria2 -y") |
|
|
|
base_url = "https://huggingface.co/camenduru/HunyuanVideo" |
|
models = { |
|
"transformers/mp_rank_00_model_states.pt": "ckpts/hunyuan-video-t2v-720p/transformers", |
|
"vae/config.json": "ckpts/hunyuan-video-t2v-720p/vae", |
|
"vae/pytorch_model.pt": "ckpts/hunyuan-video-t2v-720p/vae", |
|
"text_encoder/config.json": "ckpts/text_encoder", |
|
"text_encoder/generation_config.json": "ckpts/text_encoder", |
|
"text_encoder/model-00001-of-00004.safetensors": "ckpts/text_encoder", |
|
"text_encoder/model-00002-of-00004.safetensors": "ckpts/text_encoder", |
|
"text_encoder/model-00003-of-00004.safetensors": "ckpts/text_encoder", |
|
"text_encoder/model-00004-of-00004.safetensors": "ckpts/text_encoder", |
|
"text_encoder/model.safetensors.index.json": "ckpts/text_encoder", |
|
"text_encoder/special_tokens_map.json": "ckpts/text_encoder", |
|
"text_encoder/tokenizer.json": "ckpts/text_encoder", |
|
"text_encoder/tokenizer_config.json": "ckpts/text_encoder", |
|
} |
|
|
|
for file_path, folder in models.items(): |
|
os.makedirs(folder, exist_ok=True) |
|
command = ( |
|
f"aria2c --console-log-level=error -c -x 16 -s 16 -k 1M " |
|
f"{base_url}/resolve/main/{file_path} -d {folder} -o {os.path.basename(file_path)}" |
|
) |
|
logger.info(f"Scaricando: {file_path}") |
|
os.system(command) |
|
|
|
logger.info("Download completato.") |
|
|
|
|
|
def generate_video(prompt, video_size, video_length, infer_steps, seed): |
|
download_models() |
|
logger.info("Clonazione del repository...") |
|
os.system("git clone https://github.com/Tencent/HunyuanVideo /content/HunyuanVideo") |
|
os.chdir("/content/HunyuanVideo") |
|
|
|
save_path = "./results/generated_video.mp4" |
|
command = ( |
|
f"python sample_video.py " |
|
f"--video-size {video_size[0]} {video_size[1]} " |
|
f"--video-length {video_length} " |
|
f"--infer-steps {infer_steps} " |
|
f"--prompt '{prompt}' " |
|
f"--flow-reverse " |
|
f"--seed {seed} " |
|
f"--use-cpu-offload " |
|
f"--save-path {save_path}" |
|
) |
|
logger.info("Esecuzione del modello...") |
|
os.system(command) |
|
|
|
if os.path.exists(save_path): |
|
return save_path |
|
else: |
|
logger.error("Video non generato correttamente.") |
|
return None |
|
|
|
|
|
def infer(prompt, width, height, video_length, infer_steps, seed): |
|
video_size = (width, height) |
|
video_path = generate_video(prompt, video_size, video_length, infer_steps, seed) |
|
if video_path: |
|
return video_path |
|
return "Errore nella generazione del video." |
|
|
|
with gr.Blocks() as demo: |
|
gr.Markdown("# HunyuanVideo - Generazione di video basati su testo") |
|
|
|
with gr.Row(): |
|
with gr.Column(): |
|
prompt = gr.Textbox(label="Prompt", placeholder="Descrivi il tuo video (es. a cat is running, realistic.)") |
|
width = gr.Slider(label="Larghezza Video", minimum=360, maximum=1920, step=1, value=720) |
|
height = gr.Slider(label="Altezza Video", minimum=360, maximum=1080, step=1, value=1280) |
|
video_length = gr.Slider(label="Durata Video (frames)", minimum=10, maximum=300, step=1, value=129) |
|
infer_steps = gr.Slider(label="Passi di Inferenza", minimum=10, maximum=100, step=1, value=50) |
|
seed = gr.Slider(label="Seed", minimum=0, maximum=1000, step=1, value=0) |
|
submit_btn = gr.Button("Genera Video") |
|
with gr.Column(): |
|
output = gr.Video(label="Video Generato") |
|
|
|
submit_btn.click(infer, inputs=[prompt, width, height, video_length, infer_steps, seed], outputs=output) |
|
|
|
demo.launch() |
|
|
