CivArchive
    【SDXL】/Anime/bulldozer_BETA - v1.0
    NSFW
    Preview 5472791
    Preview 5472790
    Preview 5472786
    Preview 5472782
    Preview 5472781
    Preview 5472779
    Preview 5472785
    Preview 5472789
    Preview 5472788
    Preview 5472794
    Preview 5472795
    Preview 5472796
    Preview 5473004

    No reposting allowed (e.g., liblib, seaart)."

    The English part was generated by GPT.

    This model is only for learning and communication and may not be used for commercial purposes. The model was trained on four A40 GPUs using images generated by NAI3 and some artist images in the old version.

    learning_rate = 0.000001,min_bucket_reso = 256,max_bucket_reso = 2_048,train_batch_size=8,optimizer_type = "AdamW8bit"

    The model is capable of generating styles based on artist labels, and apologies are extended for using artists' styles in the training process. This is an issue that is challenging to avoid in model training and represents a contradiction. Out of respect for the artists' creative work, this model strictly prohibits commercial use.

    Quality word tags: (masterpiece), (best quality)

    Recommended negative embeddings: unaestheticXLv13unaestheticXL | Negative TI - v1.3 | Stable Diffusion Embedding | Civitai, negativeXL_DnegativeXL - D | Stable Diffusion Embedding | Civitai

    Negative words used by me: unaestheticXLv13, negativeXL_D, bad anatomy, blurry, disembodied limb, Two navel eyes, worst quality, low quality, More than five fingers in one hand, More than 5 toes on one foot, hand with more than 5 fingers, hand with less than 4 fingers, ad anatomy, bad hands, mutated hands and fingers, extra legs, extra arms, interlocked fingers, duplicate, cropped, text, jpeg, artifacts, signature, watermark, username, blurry, artist name, trademark, title, muscular, sd character, multiple view, Reference sheet, long body, malformed limbs, multiple breasts, cloned face, malformed, mutated, bad anatomy, disfigured, bad proportions, duplicate, bad feet, artist name, extra limbs, ugly, fused anus, text font UI, missing limb.

    Recommended style tags: [nai3], [artist mmd], [realistic, octanerender, 3DCG]

    Most artist labels have not been tested, and weights can be flexibly adjusted based on performance. Some may even require a reduction in weight when used individually:

    artist(岸田メル), game cg, Galgame, Galgame CG, artist(雨音), artist(甘城なつき), Nachoneko, realistic, octane, render, 3D CG, artist ke-ta, artist hanekoto, artist azure, artist 加瀬大輝, artist alpha, artist 键摄-铸焊锻轧冲压中, artist Hiten, artist WERI, artist roha, artist GongHa, artist 鬼针草, artist AGM, artist void0, artist mmd, artist 絆-Kizuna, artist GiO, artist kayo 6795, artist Mafuyu, artist Deyui デユイ, artist 赤隼, artist Archive, artist ミルクセーキ, artist 1 z282g, artist 1 na-ga, artist 1 リン☆ユウ, artist 荻pote, artist 萩 hagi, artist 遠坂あさ, artist Anmi, solo, artist 三渲二, artist 儒宅, nai3.

    Image output performance can be referenced from the cover image.

    禁止转载(例如:liblib,seaart)

    本模型仅用于学习交流,不得用于商业用途。该模型在旧版本上使用了4张A40显卡对NAI3生成的图片以及部分艺术家图片进行了训练。

    学习率为0.000001,最小桶分辨率为256,最大桶分辨率为2,048,训练批次大小为8,优化器类型为"AdamW8bit"。

    训练底模为Animagine XL V3

    该模型能够根据艺术家标签生成相应的风格,为了模型训练使用了艺术家的风格,对此表示歉意。这是在模型训练中难以避免的问题,也是一个矛盾之处。出于对艺术家创作的尊重,此模型严禁商业用途。

    质量词标签:(masterpiece), (best quality)

    推荐负面嵌入:unaestheticXLv13unaestheticXL | Negative TI - v1.3 | Stable Diffusion Embedding | Civitai, negativeXL_DnegativeXL - D | Stable Diffusion Embedding | Civitai

    我本人使用的负面词:unaestheticXLv13, negativeXL_D, bad anatomy, blurry, disembodied limb, Two navel eyes, worst quality, low quality, More than five fingers in one hand, More than 5 toes on one foot, hand with more than 5 fingers, hand with less than 4 fingers, ad anatomy, bad hands, mutated hands and fingers, extra legs, extra arms, interlocked fingers, duplicate, cropped, text, jpeg, artifacts, signature, watermark, username, blurry, artist name, trademark, title, muscular, sd character, multiple view, Reference sheet, long body, malformed limbs, multiple breasts, cloned face, malformed, mutated, bad anatomy, disfigured, bad proportions, duplicate, bad feet, artist name, extra limbs, ugly, fused anus, text font ui, missing limb.

    推荐的风格标签:[nai3], [artist mmd], [realistic, octanerender, 3DCG]

    艺术家标签大部分未经测试,根据表现可灵活调节权重,可以叠加使用,有些甚至单独使用都需要降低权重:

    artist(岸田メル), game cg, Galgame, Galgame CG, artist(雨音), artist(甘城なつき), Nachoneko, realistic, octane, render, 3D CG, artist ke-ta, artist hanekoto, artist azure, artist 加瀬大輝, artist alpha, artist 键摄-铸焊锻轧冲压中, artist Hiten, artist WERI, artist roha, artist GongHa, artist 鬼针草, artist AGM, artist void0, artist mmd, artist 絆-Kizuna, artist GiO, artist kayo 6795, artist Mafuyu, artist Deyui デユイ, artist 赤隼, artist Archive, artist ミルクセーキ, artist 1 z282g, artist 1 na-ga, artist 1 リン☆ユウ, artist 荻pote, artist 萩 hagi, artist 遠坂あさ, artist Anmi, solo, artist 三渲二, artist 儒宅, nai3.

    出图表现可参考封面图。

    Description

    regular model.

    FAQ

    Comments (32)

    einar_rainhartJan 14, 2024
    CivitAI

    What model is "anm3"? Animagine? Something else?

    NekoFiJan 14, 2024

    Yes, anm3 is AnimagineXL 3, and the creator of this model violates the AnimagineXL3 model licensing. AnimagineXL3 has a Fair AI Public licensing 1.0-SD license, which requires all fine tunes or merges to clearly indicate the source model and merge recipe.

    https://civitai.com/models/260267/animagine-xl-v3
    https://freedevproject.org/faipl-1.0-sd/

    einar_rainhartJan 14, 2024

    @NekoFi Thanks, as I expected. The license also requires to release derivative works under a similar license that grants the same permissions as the original. This model's permissions forbid commercial use, generation services, and in general put additional restrictions on top, and that is not allowed. No, I'm not playing police at all (honestly!), I just want to make sure people understand the terms of the models they work with.

    HansSchmidt
    Author
    Jan 14, 2024

    @einar_rainhart Hello, I am willing to abide by the rules. What text should I add to the description?

    einar_rainhartJan 14, 2024

    @HansSchmidt You'll need to:

    - Allow civitai generation (note that they will be bound by the same license terms, so they won't be able to make it private)
    - Allow selling models (ditto)
    - Tick "same permissions required" so that it shows up as such (meaning that everyone else who, says, uses this model, will be required to use the same permissions).

    I don't know if you can keep "no selling images" ticked or not, probably not to be on the safe side.

    HansSchmidt
    Author
    Jan 14, 2024

    @einar_rainhart I have selected 'No commercial use.

    einar_rainhartJan 14, 2024

    @HansSchmidt You can't. That is the crux of the matter. "No commercial use" is an additional restriction on top of the Fair AI Public License. And that is not allowed per the license itself (under "Notices"):

    "This software, all source code, and all modifications must be provided under this license or another license that allows everything this license allows. Note that this does not give you permission to change the license for this software."

    ItsMeBellJan 14, 2024
    CivitAI

    【SDXL】/Anime/bulldozer_BETA uses the AnimagineXL v3 Base Model and as stated in the license used by AnimagineXL v3, namely Fair AI Public License 1.0-SD, Models trained using AnimagineXL v3 must clearly state the base model used and use the same license as the base model, namely Fair AI Public License 1.0-SD. Modifications to AnimagineXL v3 are required to include training recipes, source code, which are not listed in the following model.
    When using the AnimagineXL 3.0 model, it's crucial to respect and adhere to the Fair AI Public License 1.0-SD to ensure ethical use and to contribute positively to the advancement of AI technology. Let's work together to uphold these principles!

    einar_rainhartJan 14, 2024

    Not only that, (writing this down for posterity) as I wrote elsewhere, you can't add additional restrictions on top of the license of the derived model: it must grant the same permissions as the original, which means you can't, for example, slap on a "no commercial use allowed".

    HansSchmidt
    Author
    Jan 14, 2024

    I am willing to comply. What kind of text should I add to the description?

    HansSchmidt
    Author
    Jan 14, 2024

    Can the current modifications be accepted?

    ItsMeBellJan 14, 2024

    Thank you for your understanding, this is for the sake of creating a healthy and educated open source environment.

    You can add training parameters that are used as in the following example.

    and also because by default Civitai uses the CreativeML Open RAIL++-M license

    you could add a clear explanation of the license used to the description like this example.

    illyaeaterJan 15, 2024

    @einar_rainhart You can't change the license on the civit upload to anything other than the options provided by this website. You can add the original huggingface license disclaimers of the model, but the website's rules will come first when it comes to its use on the website itself.

    einar_rainhartJan 15, 2024

    @illyaeater Thanks, I didn't know this. Then CivitAI must fix this, because their own rules cannot overrule the ones of the license in question (which comes from copyright law and provisions).

    bilibilissrJan 14, 2024
    CivitAI

    支持

    HansSchmidt
    Author
    Jan 27, 2024· 1 reaction

    The new version has been uploaded!

    ppkkmoonJan 14, 2024
    CivitAI

    Great !!!

    HansSchmidt
    Author
    Jan 14, 2024
    CivitAI

    I was mistakenly accused of being foolish due to an error in the translation of the description. It was my oversight, and I apologize

    ItsMeBellJan 14, 2024

    It's okay, friend, the model you made is very good, it fixes some of the shortcomings in AnimagineXL.

    we are just voicing a license that we hope can create a healthy and educated open source environment, considering that the SDXL Anime Model has been stuck for a long time.
    Thank you for considering using animagineXL as a base training model.

    I hope your model can be known so that we can provide more choices of SDXL Anime Models.

    einar_rainhartJan 14, 2024

    I think most comments were in good faith. Personally I also think you did things in good faith: licenses are complicated.

    However the sooner people understand what they can (or can't) do, the better for the ecosystem as whole.

    HansSchmidt
    Author
    Jan 14, 2024
    CivitAI

    I won't be on Civitai for a long time, and this model will be left here. Subsequent versions may not exist. If there are any errors in the description text, please forgive me. I apologize sincerely

    AlexAlexJan 20, 2024

    Will you be somewhere else or just taking a break?

    ppkkmoonJan 21, 2024

    Sad to hear that, I hope we will have a update ver.

    HansSchmidt
    Author
    Jan 24, 2024

    @AlexAlex rest

    HansSchmidt
    Author
    Jan 24, 2024

    @ppkkmoon There will be a new version next month

    HansSchmidt
    Author
    Jan 27, 2024

    @AlexAlex The new version has been uploaded!

    HansSchmidt
    Author
    Jan 27, 2024

    @ppkkmoon The new version has been uploaded!

    colaisdeliciousJan 15, 2024
    CivitAI

    I support you in the future.

    HansSchmidt
    Author
    Jan 27, 2024

    The new version has been uploaded!

    amaliarisella1906Jan 21, 2024· 2 reactions
    CivitAI

    This is probably the closest model you can use to generate similar results to NovelAI v3, amazing model

    HansSchmidt
    Author
    Jan 27, 2024

    The new version has been uploaded!

    @HansSchmidt Most anime models have this weird plastic result, so having a model that actually generate anime artstyle is amazing. Thank you for your efforts!

    Checkpoint
    SDXL 1.0

    Details

    Downloads
    1,143
    Platform
    CivitAI
    Platform Status
    Available
    Created
    1/14/2024
    Updated
    5/13/2026
    Deleted
    -

    Available On (1 platform)

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