Discord : https://discord.gg/es6NFfHtNt
Recommended KSampler :
30-100 Steps
6 CFG
Euler A
Karras
Available for freelance and commercial work | I can build custom AI models, fine-tune existing models, develop tailored AI workflows, and handle practical integrations or implementations for commercial or independent projects. Feel free to contact me if you need a custom solution.
Model Attribution & License Notice
Odali is a modified derivative of CyberRealistic Pony v18 CoreShift by Cyberdelia.
Odali is an independently developed fine-tune and is not an official release by, affiliated with, sponsored by, or endorsed by Cyberdelia.
The original base model is distributed under the CreativeML Open RAIL++-M License. Redistribution and use of Odali remain subject to the applicable terms and use restrictions of that license.
By downloading, purchasing, redistributing, or using Odali, you agree to comply with the applicable CreativeML Open RAIL++-M terms and restrictions.
Training Information
Odali was independently fine-tuned using tens of thousands of training images from a separately assembled dataset consisting of material obtained from publicly accessible sources.
The dataset was assembled and used in good faith, and no copyright infringement is intended. Public accessibility does not necessarily mean that individual materials are free of copyright or other third-party rights.
To the best of my knowledge, no material was intentionally included in violation of applicable rights. If you are a copyright holder, rights holder, creator, performer, or authorized representative and believe that material associated with Odali raises a legitimate copyright or ownership concern, please contact me with sufficient information to identify the material in question.
I am open to legitimate takedown and rights-related requests and will make reasonable efforts to investigate and accommodate valid requests to the best of my ability.
Credits
Model name: Odali
Base model: CyberRealistic Pony v18 CoreShift
Original creator: Cyberdelia
Derivative fine-tune: Odali
Training dataset: Independently assembled dataset containing tens of thousands of images
License
CreativeML Open RAIL++-M
Odali has been modified from the original CyberRealistic Pony v18 CoreShift checkpoint. Applicable attribution notices, license requirements, and use-based restrictions inherited from the original model remain in effect.
Odali is an independent derivative project and must not be represented as an official Cyberdelia or CyberRealistic release.
Description
Initial Release
FAQ
Comments (2)
Great model
Well, guys. To use most of Pony LoRAs with this checkpoint you should use "Kohya Deep Shrink"... those who can. Put that node after last (probably dmd2_4step) lora and here is the settings:
block number: 1
downscale factor: around 1.4
start at: 0
end at: 350-750 (depends on sampler settings)
That is it. Double base model resolution and enjoy the result. With those settings and dmd2_4_step_lora you may rise sampler STEPS and CFG realy high without break anythig. Like around 20 STEPS and 2.0 CFG.
Still, some LoRAs may break composition a little, if prompts is not so ovbious and/or simple. Depends on LoRA - some break, some not. Cheers.