CivArchive
    EVOKE - Semi-Realistic Fantasy Portrait Style - V1.0
    Preview 52633773Preview 52451021Preview 52686932Preview 52688541Preview 52701403Preview 52633777Preview 52633775Preview 52691231

    EVOKE - Semi-Realistic Fantasy Portrait Style v1.0

    Things To know:

    • You should use the trigger word EVOKE

    • This LoRA works very well in portrait style images!

    • I've found that sometimes eyes, weapons, or items in hands don't render as well with the SDXL version. (Flux Version Coming Soon!)

    • Occasionally places a faint edge around the image, not sure why yet but I am working on a solution.

    SDXL Usage Tips :

    • The LoRA weight I have been using is 0.85, but feel free to try and experiment!

    • When generating pictures for the examples I used Comfy UI and the settings I used are:


      LoRA: strength_model = 0.85 / strength_clip = 1.0
      Sampler: dpmpp_3m_sde
      Scheduler: Karras
      Steps: 20 - 50
      cfg: 3.5 - 7.0

    • Example of prompts used:

      "EVOKE, a semi-realistic digital painting of one solo old man, with white hair, blue eyes, and a short white beard. He is adorned with a purple cloak and silver metal armor, with a serious expression on his face. He is towards the viewer during a fading sunset in a desert landscape with the sand blowing through the air, catching the light as the wind rolls behind him with intricate details and a fantastical atmosphere. The composition is dominated by warm tones and an orange sunset with pink and dark purple gradient sky as the backdrop."

    • First image is WITHOUT my LoRA and second is WITH my LoRA, same seed, prompt, and model.

    Description

    Improved stylistic properties and artistic boundaries were sufficiently increased by a large weighted margin.

    LORA
    SDXL 1.0

    Details

    Downloads
    200
    Platform
    CivitAI
    Platform Status
    Available
    Created
    1/18/2025
    Updated
    11/12/2025
    Deleted
    -

    Files

    EVOKE_-_Semi-Realistic_Fantasy_FLUX-000037.safetensors