Nova Cross AM
Nova Cross AM is Anima checkpoint with stylized look
Each version has its own style
All images on example were created with custom diffusers-anima with custom png tags
Rules
You cannot use the generated images for commercial use if it's not edited (or just turning it to black and white)
You can share images without any restriction if you don't monetize it
Advertising this model to outside is always welcome
Recommend Settings
Sampler: Euler a Normal
Steps: 20~30
Clip Skip: 1-2
Denoising Strength: 0.65 - 0.8
CFG Scale: 4~6
Prompt: masterpiece, best quality, score_9, score_8, score_7, year 2025, newest, highres, absurdres, very aesthetic, {Prompt}
Negative Prompts: worst quality, low quality, early, old, score_1, score_2, score_3, cartoon, graphic, painting, crayon, graphite, abstract, glitch, deformed, mutated, ugly, disfigured, long body, bad anatomy, bad hands, missing fingers, extra fingers, extra digits, fewer digits, cropped, very displeasing, artist name, blurry, jpeg artifacts, lowres, censor
Description
Initial version for Eastern / Western Anime hybrid
Anima-base v1.0 based
FAQ
Comments (3)
I just can't figure out how do you need to connect CLIP Loader for it to work.
nvm I am stupid, it works
Since Nova Cross AM is based on the Anima/Qwen architecture, it cannot be loaded using the standard unified "Load Checkpoint" node. You need to load the model UNET, CLIP, and VAE separately. Here is the correct step-by-step base node setup based on a working pipeline.
Required Nodes & Connections:
Load Diffusion Model (UNET):
Select novaCrossAM_v10C.safetensors (or your current version).
Connect the MODEL output directly to the KSampler.
Load CLIP:
Select the specialized Anima/Qwen text encoder (e.g., qwen306BBaseAnima_1.safetensors.
Connect the CLIP output to both CLIP Text Encode (Positive) and CLIP Text Encode (Negative) nodes.
Load VAE:
Select the compatible VAE (e.g., qwen_image_vae.safetensors).
Connect the VAE output directly to the VAE Decode node.
Latent & Sampler:
Set Empty Latent Image to 1280x1280 (optimal for 1MP generation).
In KSampler, set cfg to 4.0 (sweet spot for this model) and steps to 20