Text-to-Image
Diffusers
Safetensors
stable-diffusion-xl
stable-diffusion-xl-diffusers
lora
photorealistic
people
cars
textures
Instructions to use Pedro21613/PS-IMAGE-1.4 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use Pedro21613/PS-IMAGE-1.4 with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("stabilityai/stable-diffusion-xl-base-1.0", dtype=torch.bfloat16, device_map="cuda") pipe.load_lora_weights("Pedro21613/PS-IMAGE-1.4") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Inference
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- Draw Things
- DiffusionBee
PS IMAGE v1.4 — SDXL (pessoas + automóveis + texturas)
LoRA sobre stabilityai/stable-diffusion-xl-base-1.0, evolução da v1.3 (que era SD 1.5).
Treinado em T4 com 900 imgs balanceadas (300 CelebA-HQ com caption + 300 Stanford Cars + 300 DTD),
768px, rank 16 / alpha 32, 600 steps, lr 5e-5 cosine, AdamW.
Validação (mesmo prompt/seed 24000, 28 steps, CFG 6.5, 768px)
- pessoa: mantém rosto, pele detalhada — igual ou levemente melhor que base SDXL
- carro branco na chuva: reflexos no asfalto mais fortes na v14
- madeira: grão/varniz preservado
- misto pessoa+carro azul: base gera pessoa de costas sem rosto; v14 gera de frente com rosto (melhora composição que a 1.3 errava)
Pastas examples/ trazem base_* vs v14_* + compare_*.
Como usar (SDXL)
import torch
from diffusers import StableDiffusionXLPipeline
from peft import PeftModel
BASE="stabilityai/stable-diffusion-xl-base-1.0"
pipe=StableDiffusionXLPipeline.from_pretrained(BASE,torch_dtype=torch.float16,variant="fp16",safety_checker=None).to("cuda")
pipe.unet=PeftModel.from_pretrained(pipe.unet,"Pedro21613/PS-IMAGE-1.4")
pipe.unet=pipe.unet.merge_and_unload()
img=pipe("ultra realistic photo of a smiling woman in a park, detailed skin texture, natural light, sharp focus, high quality",
num_inference_steps=28,guidance_scale=6.5,height=768,width=768).images[0]
Prompts que funcionam:
- pessoas:
ultra realistic photo of ..., detailed skin texture, natural light, sharp focus, high quality - carros:
ultra realistic photo of a ..., glossy paint, detailed metal, natural light, sharp focus - texturas:
macro close-up photo of ... texture, detailed surface, even lighting, sharp focus
- Downloads last month
- 7
Model tree for Pedro21613/PS-IMAGE-1.4
Base model
stabilityai/stable-diffusion-xl-base-1.0