Fluxmania - V
Palabras Clave y Etiquetas Relacionadas
Imágenes destacadas
Parámetros Recomendados
samplers
steps
cfg
vae
other models
Consejos
Use dpmpp_2m sampler with sgm_uniform schedule for Fluxmania models.
Use Clip L full fp32 model for the best performance with Fluxmania III.
Settings of CFG 3.5 and 25-30 steps are recommended.
Euler Simple sampler is recommended for Genuine model.
Aspectos Destacados de la Versión
Recommended settings : dpmpp_2m / sgm_uniform / 25 steps / flux guidance : 2.5 to 3.5.
Recommended associated lLora : Defluxion II
Version 5 compiles several models, if you ever notice the appearance of the vertical bar on the right, it is probably because you have used an existing LoRA in fluxmania V. It is recommended to significantly reduce its weight to make this artifact disappear.
Example: Fluxmania V integrates Fluxartis II, if you still want to use this LoRA, go from the recommended weight of 0.6 to 0.3
Patrocinadores del Creador
All licensing conditions for the Flux Dev model apply. You can find the details here.
To learn about the horizontal streak artifact that appears when using lora or upscaling, please see the attached link :
https://github.com/lllyasviel/stable-diffusion-webui-forge/issues/1712
Showcase images are in PNG and include metadata, prompt & workflow.
All licensing conditions for the Flux Dev model apply. You can find the details here.
Model orientation : Photography - Realism - Portraits - Artistic nude
Fluxmania III :
File 22.07 GB : unet fp16 without VAE & clip
File 11.16 GB : unet fp8 without VAE & clip
File 15.91 GB : checkpoint fp8 (not bf16) with VAE & Clip L full fp32
Settings : dpmpp_2m - sgm_uniform / cfg 3.5 / steps 25 - 30.
I recommend using this version of Clip L with the Fluxmania model
Fluxmania I & II : Unet fp8 (no VAE, no clip) / Settings : dpmpp_2m - sgm_uniform / cfg 3.5 / steps 25.
Genuine : Unet fp16 (no VAE, no clip) / Settings : Euler Simple / cfg 3.5 / steps 30.
Detalles del Modelo
Tipo de modelo
Modelo base
Versión del modelo
Hash del modelo
Creador
Discusión
Por favor log in para dejar un comentario.