CivArchive
    WonderMix - WonderMix_v2
    NSFW
    Preview 1Preview 2Preview 3Preview 4Preview 5Preview 6Preview 7Preview 8Preview 9Preview 10Preview 11Preview 12Preview 13Preview 14Preview 15Preview 16Preview 17Preview 18Preview 19Preview 20

    Merged model with additional fine-tuning for variety and quality. Midjourney-esque with the ability generate highly detailed photo-realistic images/still frames

    Like photorealism? Try my new fine-tune 'Lomostyle'

    Capabilities

    • Complex 3D renders

    • Photorealistic images

    • NSFW/SFW generations (use tags: nsfw, sfw in prompt)

    • Abstract art

    • Cartoon stills

    • Landscape photography

    • High compatibility with LoRA models

    • Designed to handle light AND complex prompts.

    Hardcore generations are also possible but, that wasn't the main focus of this model. There are much better models for this use case, unless you're looking for a more stylized/artistic hardcore generation.

    Limitations

    • Anime (Future updates will add Anime capability.)

    • Occasionally the nsfw/sfw tags don't trigger. This can usually be addressed by:

      • Moving the sfw, nsfw tags towards the beginning or end of the prompt.

      • Using attention/emphasis ((sfw)), or (nsfw:1.2)

    • Strong VAE recommendation. (this is due to fine-tuning)

    Models used in the merge

    Models used in the merge are included in the version description bellow.

    Check out my other models

    SDXL

    SD1.5

    LoRA

    Questions or Feedback?

    Visit my thread on the Unstable Diffusion Discord Server

    Description

    Initial Release

    VAE Strongly Recommended

    https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.ckpt

    In webui settings, Uncheck: Ignore selected VAE for stable diffusion checkpoints that have their own .vae.pt next to them

    Includes

    • WonderMix_v2.safetensors

    • WonderMix_v2.ckpt

    • WonderMix_v2-fp16.safetensors

    Merged Models

    Note: Many of the sample images use the bad_promptv2 textual inversion. This is not at all required but, if you want to generate identical images to the sample images you can find the textual inversion here:

    https://huggingface.co/datasets/Nerfgun3/bad_prompt/tree/main

    Other settings:

    ETA Seed Noise Delta: 31337

    ClipSkip: 1