CivArchive
    Raehoshi Anima - v1.0
    NSFW
    Preview 134128770
    Preview 134128824
    Preview 134129005
    Preview 134129156
    Preview 134128847
    Preview 134129216

    Raehoshi Anima

    Raehoshi Anima is an enhanced iteration built upon the Anima Base v1.0 architecture. This release focuses on elevating visual style, integrating extensive new concepts, and expanding character knowledge. The ultimate goal is to deliver a more polished, balanced, and visually stunning output while remaining faithful to the core strengths of the base model.

    Installation & Requirements

    Important Note: This model does not include a built-in Text Encoder or VAE. You must download these components separately to achieve the intended results.

    File Placement Guide

    • raehoshi-anima-v1.0.safetensors goes in ComfyUI/models/diffusion_models

    • qwen_3_06b_base.safetensors goes in ComfyUI/models/text_encoders

    • qwen_image_vae.safetensors goes in ComfyUI/models/vae

    For the optimal experience and the highest quality generations, we recommend the following configurations:

    • Sampler: Euler a or ER SDE

    • Schedule Type: Beta or Normal

    • Steps: 32

    • CFG Scale: 4.0 – 5.0

    • Resolution: Any resolution up to 1536 (ensure dimensions are divisible by 32)

    • Positive Prompt: masterpiece, best quality, score_7, absurdres

    • Negative Prompt: worst quality, low quality, score_1, score_2, score_3, artist name, blurry, jpeg artifacts, bad anatomy, bad hands, bad proportions, mutation, deformed, extra digits, fewer digits, missing arms, missing legs

    Prompting Tips

    • Tag Ordering: For the most consistent results, follow this structured prompt order:

      [Quality / Meta / Year / Safety tags] ➔ [1girl / 1boy / Character Count] ➔ [Character Name] ➔ [Series / Copyright] ➔ [Artist] ➔ [General Tags]

    • Character Accuracy: Always include the official series/copyright tags alongside the character name to significantly improve details and accuracy.

    • Hybrid Prompting: The model handles hybrid prompting seamlessly. Feel free to mix dan match danbooru-style tags with natural language descriptions (e.g., use tags for characters and natural language for background/action).

    Training Details

    Raehoshi Anima was trained using a custom personal fork of Diffusion-pipe across a comprehensive two-stage fine-tuning process. The dataset utilizes multi-level captioning with random selection and tag dropout to ensure flexibility.

    Stage 1: Concept & Character Expansion

    • Dataset Size: ~25k images

    • Trained Resolution: 1024x1024

    • Hardware: NVIDIA RTX PRO 6000 (96GB VRAM)

    • Batch Size: 32

    • Gradient accumulation steps: 1

    • Learning Rate: 1.5e-6 (LLM Adapter LR: 2e-7)

    • Focus: Introducing new franchises, series, and character knowledge.

    Stage 2: Aesthetic & Style Refinement

    • Dataset Size: ~1k high-curation images

    • Trained Resolution: Multi-aspect (1024x1024 & 1536x1536)

    • Hardware: NVIDIA RTX PRO 6000 (96GB VRAM)

    • Batch Size: Per-resolution batch size (24-1536x1536) & (48-1024x1024)

    • Gradient accumulation steps: 1

    • Learning Rate: 1e-6 (LLM Adapter LR: 0)

    • Focus: Mitigating artifacts, balancing composition, and enhancing the overall visual style.

    List of New Series/Characters Trained:

    Expanded Knowledge Base (Up to May 2026)

    The model’s character and lore library has been updated to include the latest data for:

    • Zenless Zone Zero

    • Wuthering Waves

    • Honkai: Star Rail

    • Genshin Impact

    • Arknights: Endfield

    • Neverness to Everness

    For character trait details prompts, please refer to the Danbooru site for accurate tags and references.

    Special Thanks

    A huge thank you to GSlinux for providing the development support needed to make this project a reality.

    Support the Development

    If you love using this model and want to help fund future iterations and dataset curation, consider supporting the project:

    • ⚡ Send a tip of Yellow Buzz directly on this platform.

    • ☕ Buy me a coffee via Ko-fi

    License

    This model is released under the CircleStone Labs Non-Commercial License.

    Description

    Initial release

    FAQ

    Comments (13)

    1254902153gxc539Jun 18, 2026· 4 reactions
    CivitAI

    Your model > World Cup final

    hushichoJun 18, 2026· 6 reactions
    CivitAI

    I've only used 1.0 a little but I have to say that it already does some of the smoothest lineart I've seen.

    Raelina
    Author
    Jun 19, 2026

    Thank you

    alicesoyJun 19, 2026
    CivitAI

    This is a model with great style, but it doesn't follow prompts very well.

    Raelina
    Author
    Jun 19, 2026· 2 reactions

    Thank you for the feedback. I will try to improve it in the future update

    ligerJun 22, 2026· 3 reactions
    CivitAI

    Kudos for Raehoshi Anima; just like Raehoshi illust XL, it is incredibly convenient and highly practical, enabling the creation of a vast number of new characters directly from the base model while delivering high-quality styling.

    Thank you!

    Raelina
    Author
    Jun 23, 2026· 1 reaction

    Thank you

    ligerJun 23, 2026

    @Raelina Hi Raelina, after conducting further tests on Raehoshi Anima V1, I would like to share a few observations with you:

    (1) For some reason, the character "cyrene \(demiurge\) \(honkai: star rail\)" doesn't look quite right—especially regarding the outfit and accessories—and using "cyrene \(ripples of past reverie\) \(honkai: star rail\)" doesn't seem to yield great results either. Given that she was previously known as "cyrene \(ripples of past reverie\) \(honkai: star rail\)" before the naming convention changed, I wonder if that shift has caused some issues.

    (2) Adherence to natural language prompts seems weaker, particularly in complex scenarios involving interactions between three people. Even when using natural language to describe the three individuals specifically via positional logic, the composition often ends up somewhat chaotic. In contrast, the same natural language prompts work well in AnimaBaseV1. Could I ask whether your current training uses a pure tagging strategy or a hybrid approach combining tags and natural language?

    These are just some personal findings for your reference.

    Raelina
    Author
    Jun 24, 2026· 1 reaction

    @liger Hi thank you so much for the detailed feedback.

    1. After double-checking my dataset, it turns out Cyrene was actually excluded from my dataset. I will make sure to include her in a future training update.

    2. My dataset uses a hybrid approach, combining both tags and natural language captions. However, I’ll explore different training strategies to improve prompt adherence issues in the future update.

    Thanks again for taking the time to test the model and share your findings.

    GoonetteAI_Jun 22, 2026· 2 reactions
    CivitAI

    It's a great model, the fact that it can actually recognise newer characters is already pretty great. The smooth and sharp lineart is pretty nice too, seems to also have less artifacts than other anima finetunes.

    I'll keep using this model for now and testing different artists but so far I'm really happy with it.

    Raelina
    Author
    Jun 23, 2026

    Thank you

    Madafada1991Jun 23, 2026· 2 reactions
    CivitAI

    Its great. Though slightly weaker on prompt side compared to other more popular Anima checkpoint. Which clearly had a whole load of improvement path to be made. But the quality of output is crazy. Sasuga Raelina-sama!

    claps in Demiurge smile

    Raelina
    Author
    Jun 24, 2026· 2 reactions

    Thank you for the feedback! I'm glad you like the quality.

    As for the prompt adherence, it might be due to my dataset structure. I'll try a different approach in the future update to fix this issue.

    Checkpoint
    Anima
    by Raelina
    Download (Beta) View on CivitAI
    anima
    base model
    game
    characters

    Details

    Downloads
    805
    Platform
    CivitAI
    Platform Status
    Available
    Created
    6/18/2026
    Updated
    6/28/2026
    Deleted
    -

    Files

    qwen_image_vae.safetensors

    Size:
    242.11 MB
    SHA256:
    50035174a66fca476700eba6c85a3e8e3e284f8c5d9026954402fbd4b1306463

    Mirrors

    CivitAI (1 mirrors)

    raehoshiAnima_v10_txt.safetensors

    Size:
    1.11 GB
    SHA256:
    cd2a512003e2f9f3cd3c32a9c3573f820bb28c940f73c57b1ddaa983d9223eba

    Mirrors

    HuggingFace (71 mirrors)
    CivitAI (66 mirrors)
    ModelScope CN (2 mirrors)

    raehoshiAnima_v10.safetensors

    Size:
    3.90 GB
    SHA256:
    146f9197153d1ce5dac680f23fae940e0ce94857fd86504868df23692721be2d

    Mirrors

    HuggingFace (1 mirrors)
    CivitAI (1 mirrors)