|
import gradio as gr |
|
import spaces |
|
import os |
|
import torch |
|
from datetime import datetime |
|
from PIL import Image |
|
import boto3 |
|
from botocore.exceptions import NoCredentialsError |
|
from dotenv import load_dotenv |
|
from diffusers import DiffusionPipeline, EulerAncestralDiscreteScheduler, EulerDiscreteScheduler |
|
|
|
|
|
load_dotenv() |
|
|
|
|
|
AWS_ACCESS_KEY = os.getenv('AWS_ACCESS_KEY') |
|
AWS_SECRET_KEY = os.getenv('AWS_SECRET_KEY') |
|
AWS_BUCKET_NAME = os.getenv('AWS_BUCKET_NAME') |
|
AWS_REGION = os.getenv('AWS_REGION') |
|
HF_TOKEN = os.getenv('HF_TOKEN') |
|
|
|
|
|
s3_client = boto3.client( |
|
's3', |
|
aws_access_key_id=AWS_ACCESS_KEY, |
|
aws_secret_access_key=AWS_SECRET_KEY, |
|
region_name=AWS_REGION |
|
) |
|
|
|
|
|
character_pipe = DiffusionPipeline.from_pretrained( |
|
"cagliostrolab/animagine-xl-3.1", |
|
torch_dtype=torch.float16, |
|
use_safetensors=True, |
|
use_auth_token=HF_TOKEN |
|
) |
|
character_pipe.scheduler = EulerDiscreteScheduler.from_config(character_pipe.scheduler.config) |
|
|
|
|
|
item_pipe = DiffusionPipeline.from_pretrained( |
|
"openart-custom/DynaVisionXL", |
|
torch_dtype=torch.float16, |
|
use_safetensors=True, |
|
use_auth_token=HF_TOKEN |
|
) |
|
item_pipe.scheduler = EulerAncestralDiscreteScheduler.from_config(item_pipe.scheduler.config) |
|
|
|
|
|
@spaces.GPU(duration=60) |
|
def generate_image(model_type, prompt, negative_prompt, width, height, guidance_scale, num_inference_steps): |
|
if model_type == "character": |
|
pipe = character_pipe |
|
default_prompt = "1girl, souji okita, fate series, solo, upper body, bedroom, night, seducing, (sexy clothes)" |
|
default_negative_prompt = ("lowres, (bad), text, error, fewer, extra, missing, worst quality, jpeg artifacts, " |
|
"low quality, watermark, unfinished, displeasing, oldest, early, chromatic aberration, " |
|
"signature, extra digits, artistic error, username, scan, [abstract]") |
|
elif model_type == "item": |
|
pipe = item_pipe |
|
default_prompt = "great sword, runes on blade, acid on blade, weapon, (((item)))" |
|
default_negative_prompt = "1girl, girl, man, boy, 1man, men, girls" |
|
else: |
|
return "Tipo inválido. Escolha entre 'character' ou 'item'." |
|
|
|
|
|
final_prompt = prompt if prompt else default_prompt |
|
final_negative_prompt = negative_prompt if negative_prompt else default_negative_prompt |
|
|
|
|
|
pipe.to("cuda") |
|
|
|
|
|
result = pipe( |
|
prompt=final_prompt, |
|
negative_prompt=final_negative_prompt, |
|
width=int(width), |
|
height=int(height), |
|
guidance_scale=float(guidance_scale), |
|
num_inference_steps=int(num_inference_steps) |
|
) |
|
image = result.images[0] |
|
|
|
|
|
temp_file = "/tmp/generated_image.png" |
|
image.save(temp_file) |
|
|
|
|
|
file_name = datetime.now().strftime("%Y%m%d_%H%M%S") + ".png" |
|
try: |
|
s3_client.upload_file(temp_file, AWS_BUCKET_NAME, file_name) |
|
s3_url = f"https://{AWS_BUCKET_NAME}.s3.{AWS_REGION}.amazonaws.com/{file_name}" |
|
return s3_url |
|
except NoCredentialsError: |
|
return "Credenciais não disponíveis" |
|
|
|
|
|
def gradio_generate(model_type, prompt, negative_prompt, width, height, guidance_scale, num_inference_steps): |
|
return generate_image(model_type, prompt, negative_prompt, width, height, guidance_scale, num_inference_steps) |
|
|
|
|
|
model_type_input = gr.Dropdown(choices=["character", "item"], value="character", label="Model Type") |
|
prompt_input = gr.Textbox(lines=2, placeholder="Digite o prompt (deixe vazio para o padrão)", label="Prompt") |
|
negative_prompt_input = gr.Textbox(lines=2, placeholder="Digite o negative prompt (deixe vazio para o padrão)", label="Negative Prompt") |
|
width_input = gr.Number(value=832, label="Width") |
|
height_input = gr.Number(value=1216, label="Height") |
|
guidance_scale_input = gr.Number(value=10.0, label="Guidance Scale") |
|
num_inference_steps_input = gr.Number(value=100, label="Number of Inference Steps") |
|
|
|
|
|
iface = gr.Interface( |
|
fn=gradio_generate, |
|
inputs=[ |
|
model_type_input, |
|
prompt_input, |
|
negative_prompt_input, |
|
width_input, |
|
height_input, |
|
guidance_scale_input, |
|
num_inference_steps_input, |
|
], |
|
outputs="text", |
|
title="Image Generation API", |
|
description="Gere imagens usando modelos de difusão e faça upload para o AWS S3." |
|
) |
|
|
|
if __name__ == "__main__": |
|
iface.launch() |
|
|