Photomatix v2 Changes

• experimental model merge aimed at more cinematic outputs

• tuned to work with Photomatix and SDXL styles: Download the full collection of styles from Github: PX-Styles.

• standard resolutions (512x512, 512x768, 768x768, 960x640, 1024x768, 768x896, 1024x1200)

Check the gallery for examples. Read more about the v2 model (link to the article)

RT (Experimental) Version: Tested on A4000 (NOT tested on other RTX Ampere cards, such as RTX 3090 and RTX A6000). This version is for advanced users and needs TensorRT extension and NVIDIA TensorRT.
RT（实验）版本：在A4000上测试（未在其他RTX Ampere卡（如RTX 3090和RTX A6000）上测试）。此版本适用于高级用户，需要TensorRT扩展和NVIDIA TensorRT。

Read about the installation carefully in this article. This version doubles the render speed with a maximum working size of 832x832. After the installations, download the .trt file (hosted on Hugginface) into the stable-diffusion-webui\models\Unet-trt. Then you can choose it in the GUI list as in the tutorial. I recommend creating another installation of A1111 for these experiments, as described in the article.

Photomatix is the merge I use for my experiments. I am uploading it for consistency with my other models. It performs surprisingly well in many resolutions and setups and the results are comparable to the best merged and trained base models. It can create a photorealistic output but also various creative styles for concept design. Read more about the model and its use here.
Photomatix是我用来做实验的合并工具。我上传它与我的其他模型的一致性。它在许多分辨率和设置中表现得令人惊讶，结果与最佳合并和训练的基础模型相当。它可以创建逼真的输出，也可以为概念设计创建各种创意风格。阅读更多关于模型及其使用这里.

Strong points: Consistent, universal, realistic, balanced, and flexible. Works well with extensions. Works great with LoRAs and styles. Simple negative prompt.

Weak points: Hands and feet (sometimes), Characters handling objects

Extensions to use: ControlNet, ADetailer, Regional Prompter, Latent Couple, Roop, Dynamic Thresholding, Latent Mirroring, Hires fix. (in txt2img)

Resolutions for portraits: can create quite consistent outputs f.i. for 768x512, 640x832, 768x896, 1024x768, 1112x768
肖像的分辨率：可以产生相当一致的输出，适用于768x512、640x832、768x896、1024x768、1112x768

Recommended CLIP SKIP: 1-2

There is the inpainting version available in the menu at the top of the page. Although you can inpaint with the normal version too (and sometimes the result is quite good), you may see the differences in the gallery. The inpaint version is just better for big changes.

Quick Tutorial on Inpainting Details in A1111:

• Load your image (PNG Info tab in A1111) and Send to inpaint, or drag and drop it directly in img2img/Inpaint

• create or modify the prompt as needed (if you have imported an image with generation data, do not forget to uncheck the former base model, if it is connected!)

• Load the inpainting version of Photomatix model

• Mask the area you want to change and Generate

If you want to learn more about AI-augmented design and algorithmic art, check my web pages.

For my notes on training, research, and use of stable diffusion models, check https://sandner.art/articles/ai/

If you can afford it, support my work by buying me a coffee to keep those costly render cores going and to keep the creative flow: https://ko-fi.com/airesearch

If you can not, review and rate the model to help others to find it.

Have fun rendering!

Checkpoints used for v1 merge:

• https://civitai.com/models/39044/c3

• https://civitai.com/models/49934/crystal-clear2

• https://huggingface.co/runwayml/stable-diffusion-inpainting

• https://huggingface.co/runwayml/stable-diffusion-v1-5/tree/main

Checkpoints used for v2 merge:

• Photomatix v1

• Photon v1 (Civitai link) 

• AbsoluteReality (Civitai link) 

• epiCRealism (Civitai link) was also used to balance the refiner versions 

Thanks to the community and contributors for their hard work, without the collective effort and resources such projects would not be possible.