Azure & Crimson Anime-Centric Merged Model
📌 Model Type: Merged Checkpoint
🧠 Base Components Merge: V1 - V2, V4-V5
Illustrious XL 2.0
WAI-NSFW-Illustrious-SDXL
Nova Furry XL
Absolute Territory
I previously mentioned that V4 Final would be the last update, but after receiving multiple requests in PM, I decided to continue development and release this version Both V5 and V4 will be up for generation onsite.
V5 Update
This Version act as stability update and build in Additional character/style up to Early March 2026
🔧 Stability & Performance Update
This version focuses on overall stability improvements
Improved handling of recognized characters and style consistency
Smoother generation results compared to previous versions
Faster generation performance
🧠 Model Composition
Merge and base composition remain unchanged from V4 Final
New Project Announcement Azure Starfire
Hello everyone!
Following the conclusion of the Azure & Crimson mix series, I’m excited to announce the start of a brand new project: Azure Starfire.
This new model line represents a fresh beginning built from the ground up with a clean dataset, improved balance, and updated methods based on everything learned from previous versions
🧬 Model Focus
Azure & Crimson is a merged model built on a strong foundation of anime style generation, with a particular emphasis on character centric composition and minimal background detail. It delivers vibrant colors, sharp features, and versatile rendering, making it ideal for a variety of themes.
While the model excels in anime aesthetics, it is not recommended for semi-realistic or realistic styles.
More Details in here : https://civarchive.com/articles/16158/about-azure-and-crimson-checkpoint
⚙️ Recommended Settings
Steps: 25–40
CFG Scale: 4–8
Note: Values above 9 may cause style melding or loss of detail, depending on prompt structure.
Sampler: Euler a / DPM++ 2M Karras (or user preference)
Prompting:
Works best with positive quality tags and clear structure.
Hires Fix: ✅ Recommended
FaceDetailer (ComfyUI): Optional for close-up portraits
🏷️ Prompt Tags
✅ Positive Tags (Quality Boost):masterpiece, amazing quality, best quality, ultra-detailed, 8K, illustration, CG, ultra-detailed-eyes, detailed background, very awa, newest, highres, absurdres
🚫 Negative Tags (Suggested):lowres, (worst quality:1.2), (bad quality:1.2), (low quality:1.2), bad anatomy, old
🔒 License
This model is released under the Fair AI Public License 1.0-SD, allowing for ethical and fair use in accordance with AI content creation standards.
Description
Version 4 A Test Update Changelog
Test Build Notice
V4 A is an experimental test model created to evaluate new training methods.
This build is not a full release, but serves as a testing ground for future updates.
Training Method
Introduced DreamBooth for custom data integration, replacing the earlier LoRa-based approach.
Style Integration
Integrated all custom styles developed so far.Trigger words remain the same check My profile for the full style reference.A complete Data list will be released on Discord and CivitAi Article in the final model variant.
FAQ
Comments (6)
Hi everyone We’re currently testing DreamBooth for model finetuning. It probably won’t be perfect on the first attempt, but we’d love to hear your feedback on how it turns out. This version will be released as V4 A Fine, Tomorrow
Seems like the details were lessened in the V4 A Fine
@Dimaray I think V4 A Fine requires a more specific prompt. I’ll upload the test image in a bit. From the two dragon images on the model cover, you can see that V4 A Fine is actually more accurate in terms of prompt response and ty for the feedback
@BastionExa Hmm ok, will have to test more to see if I can get it to work for me.
@BastionExa Yeah ok see what you mean, time to change my process for the CP ^^