Chroma, OmniA, V-Pred & More
This announcement was made on June 13, 2025
OmniA: The All-In-One Multi-Style Model
We’re thrilled to introduce OmniA, a groundbreaking multi-style model created by our partner @Samael1976. Like Pony and Illustrious, OmniA is designed as a next-gen base model — perfect for further training. It blends multiple artistic styles at its core, making it incredibly versatile and robust.
Play the Model: https://www.mage.space/play/258f112e19234c848e356606ea4ac545
Play Realistic Style Concept: https://www.mage.space/play/9f12e95feb544f078ce3d5feebd83c61
Play 2.5D Concept: https://www.mage.space/play/fc5687baf48c47d1bb9827542ecdcc67
Play Anime Concept: https://www.mage.space/play/409d98a5d4bc42aeb045da7ad620dbb3
Play Western Comics Concept: https://www.mage.space/play/f14e47698b044979b08ff1973bda906d
See the detailed technical overview and usage guide PDF here: https://drive.google.com/file/d/13U-thnEeIsm7r2r7F55FVmfFnolKFyTb/view?usp=sharing
V-Pred: Higher Quality with Smarter Learning
We’re introducing two exclusive V-Pred-based models by @CF— CFV Prodigy and CFV Prominence. The v-prediction method helps the model learn more efficiently by predicting both the noise to remove and the final image. This balanced training approach leads to sharper and more visually refined results.
Play CFV Prodigy: https://www.mage.space/play/f7e52874cd0c43b4aecccc779d6b3da8
Play CFV Prominence: https://www.mage.space/play/eac355ba3ce546a58559c9f2c83bdc3d
As a Pro Plus member, you can also import and use any V-Pred model — make sure to select a V-Pred scheduler when generating!
Chroma: Pony for Flux [Beta]
We are excited to bring Chroma, a new beta model architecture. Trained on a 5M dataset including anime, furry, artistic stuff, and photos, Chroma brings back missing anatomical concepts with stunning clarity. Simply put, Chroma is Pony for Flux.
Where to find: Advanced → Select Model Architecture → Chroma
Learn more: https://huggingface.co/lodestones/Chroma
See examples: https://civitai.com/models/1330309?modelVersionId=1884429
This model is still in beta (v0.36) and under active development. Results may vary — please use with care! Give us feedback, and we will make sure to share it with the model creator.
Style Reference for SD 3.5 Large
We’ve added Style Reference support for SD 3.5 Large — great for generating more stylized and artistic creations. Give it a try and let us know what you think!
Two Refine Modes Now Available
Fast Refine (V2)
Slow Refine (V3)
Selfie Is Back [Beta]
Create fun AI images from your selfie!
Available under Advanced → Architecture → Selfie
Beta feature for Pro and Pro Plus