V2.0
Spanish:
He usado una nueva configuración para que el resultado sea mucho más realista. Trata de usar EpicRealism.
Es posible que el nombre del LORA cambie, así que no lo copies directamente de CIVITAI, mejor selecciónalo del apartado de LORAS de Stable Diffusion para que no haya errores en la generación de imágenes.
Mi configuración usando EpicRealism:
Lora Strength: 1
Prompt: Explicación simple de la imagen (pruebe primero sin palabras clave adicionales)
En el caso de esta Lora, el prompt debe comenzar con "A woman...".
Prompt negativo: cartoon, painting, illustration, (worst quality, low quality, normal quality:2), asian, chinese, (epiCNegative:0.9)
Sampling steps: >20 (si la imagen tiene errores o artefactos, use steps más altos)
CFG Scale: 5
Sampling method: DPM++ SDE Karras
Tamaño: 512x768 o 768x512
Hires fix: 4x_NMKD-Superscale-SP_178000_G (Denoising strenght: 0.35, Upscale by: 2)
English:
I have used a new configuration so that the result is much more realistic. Try using EpicRealism.
It is possible that the name of the LORA will change, so do not copy it directly from CIVITAI, rather select it from the LORAS section of Stable Diffusion so that there are no errors in the generation of images.
My setup using EpicRealism:
Lora Strength: 1
Prompt: simple explanation of the image (try first without extra keywords)
In the case of this LORA, the prompt must begin with "A woman...".
Negative prompt: cartoon, painting, illustration, (worst quality, low quality, normal quality:2), asian, chinese, (epiCNegative:0.9)
Sampling steps: >20 (if image has errors or artefacts use higher Steps)
CFG Scale: 5
Sampling method: DPM++ SDE Karras
Size: 512x768 or 768x512
Hires fix: 4x_NMKD-Superscale-SP_178000_G (Denoising strength: 0.35, Upscale by: 2)
V1.0
Lora Strength: 0.8
This LORA has been trained with the Photon model.
Follow the indications of the model to have a good result:
Prompt: A simple sentence in natural language describing the image.
In the case of this Lora, the prompt must begin with "a woman...". Better separately: "A woman, (the rest of the indications),...
Negative: "cartoon, painting, illustration, (worst quality, low quality, normal quality:2)"
Sampler: DPM++ 2M Karras | Steps: 25 | CFG Scale: 6
Clip Skip 2
Size: 512x768 or 768x512
Hires.fix: 4x-UltraSharp | steps: 10 | Denoising: 0.45 | Upscale x 2
(avoid using negative embeddings unless absolutely necessary)
Description
Spanish:
He usado una nueva configuración para que el resultado sea mucho más realista. Trata de usar modelos realistas como Realistic Vision o EpicRealism. En mi caso he usado EpicRealism para hacer mis imágenes, pero puede usar el que usted prefiera, siempre usando la mejor configuración de cada modelo para tener los mejores resultados. En la descripción os dejaré la configuración que he usado para hacer mis imágenes.
English:
I have used a new configuration so that the result is much more realistic. Try to use realistic models like Realistic Vision or EpicRealism. In my case I have used EpicRealism to make my images, but you can use the one you prefer, always using the best configuration of each model to have the best results. In the description I will leave you the configuration that I have used to make my images.