CivArchive
    EpicCeleb - 0.75x
    NSFW
    Preview 7899052
    Preview 7899088
    Preview 7899087
    Preview 7899056

    About The Model

    ABOUT

    I have merged the EpiCRealism and the Famous checkpoint models together. The goal of this model is to produce photorealistic results as well as having be able to take notice of the real people in your prompts. I want this model to have photorealism, prompt recognition, and celebrity/famous people detection (not sure there's a word for that).

    PROS

    1. This model is flexible with all kinds of prompts. This means that the model is flexible in recognizing a prompt without the user changing its weight.

    2. This model is really good with producing photorealistic products.

    3. Very good with clothing, except folds. Sometimes when there is a fold in a piece of clothing that a subject will wear, it will connect and fuse with the body of the subject.

    4. Insanely great at producing celebrities. No need for LoRA after seeing this model for the first time for me. It can create a mass number of existing celebrities faces.

    My Settings

    SETTINGS (recommended)

    Clip Skip (Do not change no matter what): 1

    Sampling Method: DPM++ SDE Karras. If you would like more speed then use DPM++ 2M Karras.

    Sampling Steps: 20+

    Size: 512x768

    CFG Scale: 7 or below

    PROCESS

    1. Generate image

    2. Send to Img2Img tab

    3. Set denoising strength to 0.5 and switch resize to to resize by and set that to 2.

    4. Send that finished image to the Extras tab and select ESRGAN_4x for upscaler 1

    5. Hit generate and then you should have something like I do

    Description

    0.75x of Famous was added to EpiCRealism

    FAQ

    Checkpoint
    SD 1.5

    Details

    Downloads
    277
    Platform
    CivitAI
    Platform Status
    Available
    Created
    3/13/2024
    Updated
    4/22/2026
    Deleted
    -

    Files

    epicceleb_075x.safetensors

    Mirrors

    CivitAI (1 mirrors)

    Available On (1 platform)

    Same model published on other platforms. May have additional downloads or version variants.