Type | |
Stats | 751 |
Reviews | (116) |
Published | Feb 14, 2024 |
Base Model | |
Usage Tips | Clip Skip: 2 |
Hash | AutoV2 8CEE5F642E |
Pruned Version Will Be Available After Testing
Aardvark 2024 Photography is a photorealistic model that leverages and attempts to improve on the Laion Aesthetic dataset tags and labels.
This is my second model shared with the Civitai community, I'm eager to hear your feedback. This model began as an incremental improvement to another model that I created here called "Fusioncore." - while training this model, I found that it had left the boundaries of what I was hoping to achieve with that model, and it became its own thing.
If you want to buy me a coffee, thank you! If not, that's ok too.
Donations: https://www.buymeacoffee.com/quadpipe
Some of our fellow members inspired me, thank you for your awesome contributions to the community! I used images generated from some of their models occasionally to build my training data. These are the models that helped me create the data I needed.
Analog Diffusion by wavymulder
CyberRealistic by Cyberdelia
Analog Madness by CornmeisterNL
A-Zovya Photoreal by Zovya
epiCRealism by epinikion
Juggernaut by KandooAI
This is a Stable Diffusion 1.5 model. I used the v1-5-pruned.ckpt as a starting point.
Over the last year, dozens of independent models were trained and merged with the v1-5-pruned.ckpt in order to achieve an initial v0.1 which was then incrementally improved on over the course of a year with individualized concept training through a mixture of LoRA merges and concept ckpts. This is the unpruned version, and a pruned version will be available in the coming weeks once I've tested it to ensure that there's no degradation.
I've recently been made aware that there are some users here that want to try my models but are only able to download from huggingface, I've uploaded this model there as well and will maintain both repos.
Huggingface Download: https://huggingface.co/QuadPipe/Aardvark_2024_Photography
License: https://huggingface.co/QuadPipe/Aardvark_2024_Photography/blob/main/LICENSE
Bias, Risks, and Limitations
Because this model uses the common v1.5-pruned.ckpt as a starting point, any biases, risks, or limitations inherent in that model will likely be observed with this model - perhaps to a lesser degree. Be mindful that this model is capable of NSFW content, and appropriate plugins or setting adjustments may be necessary to limit inappropriate outputs.
Recommendations
Aardvark 2024 responds well to story-based prompting and tag based prompting. Negative prompting is useful for fine-tuning. I find that negative TIs are a bit heavy, and I'd recommend only adding negative prompts for issues you are actively trying to solve.
Uses
It IS acceptable to use images generated from this model to train other models.
It IS acceptable to use images generated from this model for commercial purposes (with attribution of: "Images generated by Aardvark 2024 Photographic AI")
It IS acceptable to use images generated from this model for non-profit purposes (with attribution of: "Images generated by Aardvark 2024 Photographic AI")
It IS acceptable to use images generated from this model for Media & News organization purposes (with attribution of: "Images generated by Aardvark 2024 Photographic AI")
It IS acceptable to use images generated from this model for research purposes (Please let us know, so we can see if we can support your research by answering questions you may have)
It IS acceptable to download and install this model onto your computer or server
It IS acceptable to provide a copy of this model on a corporate, research, or educational institution intranet to allow for safe and convenient use by members.
Out-of-Scope Use
It IS NOT acceptable to use images generated from this model to simulate any living person without their permission.
It IS NOT acceptable to merge this model with other models, without my express and written permission.
It IS NOT acceptable to sell or use this model as part of a subscription service without my express written permission.
It IS NOT acceptable to reverse engineer this model.
It IS NOT acceptable to use images generated from this model for unethical purposes, which may include but not be limited to harassment, impersonation, and unauthorized product endorsements.