Download
1 variant available
fp16 SafeTensor
MASTER_ANIMA_V2_B10_V1-009172612full.safetensors
Half precision, best balance • 4.13 GB
Verified: a month ago
Qwen Image VAE • qwen_image_vae.safetensors
5670 1 2 3 4 5 6 7 8 90 1 2 3 4 5 6 7 8 90 1 2 3 4 5 6 7 8 9
1.6K0 1 2 3 4 5 6 7 8 9.0 1 2 3 4 5 6 7 8 9K
(51)
Jul 2, 2026


5.1K0 1 2 3 4 5 6 7 8 9.0 1 2 3 4 5 6 7 8 9K
47K0 1 2 3 4 5 6 7 8 90 1 2 3 4 5 6 7 8 9K
337.4K0 1 2 3 4 5 6 7 8 90 1 2 3 4 5 6 7 8 90 1 2 3 4 5 6 7 8 9.0 1 2 3 4 5 6 7 8 9K

License:
AnimaThe Anima Model is licensed by CircleStone Labs LLC. Copyright CircleStone Labs LLC. IN NO EVENT SHALL CIRCLESTONE LABS LLC BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH USE OF THIS MODEL.
Built on NVIDIA Cosmos
MASTER_ANIMA: V2 | AniMaster
Welcome to the next evolution of the Anima merging workflow. This checkpoint represents a refined, highly optimized fusion, designed to deliver the speed and simplicity of classic SD 1.5 merges, but with the stunning quality and flexibility of the modern DiT (Diffusion Transformer) architecture.
📖 Looking for the complete tutorial and the scripts used to create this?
I have moved the entire guide, including the Low-RAM Colab merging toolkit, to a dedicated article. If you want to learn how to build your own master merges in minutes, check it out here: 🔗 Mastering Anima Merges_v2
About This Version
I’ve been actively developing this ecosystem, and this release serves as a direct, highly optimized "close relative" of the original models. It proves that complex DiT merges don't have to be slow or resource-intensive. In fact, it offers Turbo-like flexibility without sacrificing the depth of a full-step model.
Key Highlights
Turbo-Like Speed & Flexibility: Works beautifully in as few as 6 steps with low CFG, but gracefully scales up to high step counts and high CFGs for maximum detail.
DiT Native: Built to respect the modular nature of Anima while providing a unified, plug-and-play DiT checkpoint.
Extreme Resolution Scaling: Thanks to the DiT architecture, this model thrives at high resolutions. It handles 5000x5000+ effortlessly without breaking or degrading (higher resolution = better quality).
Highly Versatile: An excellent base for direct generation, or as a foundation for your own custom merges and specialized finetunes.
Recommended Settings (Highly Flexible)
This model is incredibly forgiving. Adjust based on your desired speed vs. detail trade-off:
Samplers:
Euler,Euler SDE,DPM++ 2SDE Karras(Highly recommended for best results)Steps:
Fast/Turbo mode: 6 - 10 steps
High Detail mode: 20 - 50+ steps
CFG Scale:
Soft/Turbo look: 1.5 - 3.0
Standard/High Detail: 4.0 - 7.0+
Resolution: 1024x1024 (Base) up to 5000x5000+ (The higher, the better the quality and coherence).
I hope you enjoy this new iteration as much as I enjoyed building the tools behind it. If you create something amazing (especially at high resolutions!), feel free to share it in the comments.
Happy generating!


