Refiner is unnecessary and VAE is included (sdxl_vae.safetensors)
Settings for Local generations
dpmpp_3m_sde + exponential | dpmpp_sde + Karras | Euler + Normal
Steps from 20 to 40 and CFG between 3 and 7
Upscaler: 16xPSNR with Heun + Normal.
Settings for Civitai generations
DDIM or Euler | Heun or DPM2 | DPM++ or LMS
Steps from 30 to 50 and CFG between 5 and 9
Potographic Negatives: (high contrast, over saturated, glossy:1.1), (cartoon, 3d, 3d render, Photoshop, sketch, sketches, video game, draw, paint, painting, render, cgi, computer graphics, anime, manga, 2d art, 3d art, illustration:1.1), (canvas frame, watermark, signature, username, artist name:1.1),
NewRealityXL is a sophisticated merge of SDXL-based models and LoRAs that aims to be a benchmark for my experimentation whose objective is to be as consistent as possible in the generation of photographic and photorealistic contents while maintaining a high level of creativity and imagination in the generation of the subject and the style of the image.
Why chose NewRealityXL?
With an emphasis on versatility and quality, NewRealityXL offers an impressive range of photorealistic images on a wide range of subjects. From intricate portraits and expansive landscapes to imaginative depictions of fantastical creatures and anthropomorphic beings, through depictions of real people, animals, vehicles and more, all with deft attention to detail and lighting, the model excels at generating content that reflects the complexity and diversity of the real world.
How to Use NewRealityXL?
NewRealityXL is accessible through multiple channels.
Users can Download the checkpoint from the CivitAi page for local use, integrating it seamlessly with popular web interfaces such as ComfyUI or automatic1111, which provide intuitive controls and customization options.
Alternatively, for those who prefer instant access without the need for local setup, NewRealityXL is also available online on TensorArt, offering a convenient platform for generating high-quality images.
Where is aiming NewRealityXL?
The main goal of NewRealityXL is to achieve a level of image generation that is indistinguishable from real photography, eliminating common pitfalls such as artifacts, anatomical inaccuracies and other inconsistencies.
The ongoing development of NewRealityXL is driven by a commitment to excellence and continuous improvement, with a focus on refining the model to meet and exceed user expectations for realism, diversity and artistic potential.
Future update of NewRealityXL?
The focus will be on enhancing the model's precision in generating anatomical details, particularly when dealing with small pixel areas. This includes improving the representation of hands and faces in full-body images, as well as eyes or genitals in close-up views. The goal is to address and refine the model's ability to accurately render these complex anatomical features, ensuring that even in images where such details occupy a small portion of the pixel space, they are depicted with high fidelity and realism.
I invite users, artists and creators of all backgrounds to explore its capabilities and contribute to its evolution. As we push the boundaries of what is possible in image generation, your feedback and creative explorations are invaluable in shaping the future of photorealistic images.
During my experiments I tested and used hundreds of checkpoints and LoRAs, some ended up in versions that you can experiment with, others abandoned forever and others still unused but highly appreciated and inspirational.
Here I want to thank all the creators of the models that satisfied me the most, regardless of whether they were used for my purpose or not, and I invite you to give them support for the work they have done.