CivArchive
    Pastel Dreams - v0.1a Pastel Punk Anime
    NSFW
    Preview 111442
    Preview 111274
    Preview 106560
    Preview 106561
    Preview 111426
    Preview 106559
    Preview 106623
    Preview 106550
    Preview 106557
    Preview 106556
    Preview 106555
    Preview 106554
    Preview 111425
    Preview 106562
    Preview 106552
    Preview 107690

    The model is still in beta and the base model (version "b") produces a semi-realistic 3D look heavily influenced by anime.

    It's great for any RPG style and has vibrant Pastel colors.

    Additionally, if you want more of a 2D look there is an Anime optimized model, you can use the Anime optimized model (version "a") or adjust the level of anime-style rendering of either model using any LORA or hypernet between 0.05 to 0.5, depending on the style you prefer. Many of the examples were created using Wano LORA. To achieve this desired effect, install "Wanostyle" and add it to the prompt.

    https://civarchive.com/models/4219

    Please note that the examples have been created using a combination of clip skip 1 & 2.

    I don't require any credit if you use my model, but it would be greatly appreciated if you could include any of the hashtags #PastelPunkDream #PastelPunkRPGDream #PastelPunkAnime in your posts. This way, I can see what people are making with the model. Additionally, consider leaving a review so I can get a better understanding of how people are using the model and how I can improve it.

    Description

    VAE baked in, Beta Anime model.

    This model is optimized for anime, no need to use a Hypernet or LORA to get Anime results. The anime optimized model is the result of combining five different models with specific weights: 55.85% at Anything v3, 24.46875% at Synthpunk Search raw, 11.25% at DreamShaper v3.3.2, 5.625% at RPG v4, and 2.8125% at Pastel Anime mix.

    Include "(80s-90s retro anime or 20XX anime + style "name" anime:1.3)" in the prompt, depending on the desired era. You can adjust the value in the brackets, within the range of 1.05 to 1.6, to achieve a more 2D appearance and reduce the 3D cel shading effect produced by the model. With 60% of its training focused on anime, this model is highly responsive to trigger words. The use of LORA or Hypernet is optional, but they may offer more consistent results or simplify the process of creating a simple style. Additionally, adding "(3D model!:1.x)" to the negative prompt with weight can help.

    FAQ

    Checkpoint
    SD 1.5
    by Aylo

    Details

    Downloads
    177
    Platform
    CivitAI
    Platform Status
    Available
    Created
    2/16/2023
    Updated
    5/13/2026
    Deleted
    -
    Trigger Words:
    90s retro anime style
    xx year anime style

    Files

    pastelDreams_v01aPastelPunkAnime.safetensors

    pastelDreams_v01aPastelPunkAnime.ckpt

    Mirrors