Bring your own model - Walkthrough

Finetuned models in the open source AI community

What is so special about the open-source community? Just like Wikipedia, everything is openly accessible and built by a large number of people motivated to give their knowledge and experience to a greater cause. They share their codes, programs, and also their models. But what is a model? Let’s have a look at the picture with the tree again:

An overview on how AI works

Every image generation has a base model from where the prompt takes its information on what an image has to look like. It’s the source where usually billions of images have been trained. The downfall to this is that it is nearly impossible to access all of the possibilities of this model only by using words. And that is where fine-tuned models come into place.

The open-source community has a vast variety of fine-tuned models which they share on various platforms. One of them is civitai.com. You can find basically every imaginable style or object under thousands of models. This is what differentiates open source from closed source, like Midjourney. You are able to break the pattern of jumping into the same image pool as everyone else by using specific models. And you can use them directly on cogniwerk.ai without even downloading them.

All you have to do is copy the URL of the model that you found on civitai.com and paste it into the text bar under “imported models”. And seconds later it is ready to use for you inside the known Cogniwerk interface.

In pictures that is:

Step 1: Go to the tab on the left side “imported models”. Click “Import new model” and this image will appear

Step 2: Go to civitai.com and find a good model. Currently, we only provide fine-tuned LoRA models. It’s the easiest and most flexible way to use them. So best is to filter the search to a LoRA on 1.5 or SDXL as seen in the screenshot left.

Once you found a good one, copy the URL and paste it into cogniwerk.ai. Click “Import model” and a few moments later it should show in the interface ready to use for you. If it doesn’t show immediately, please try one more time. Usually, on the second try, it should be there.

If you couldn’t find the specific style you want, try to train a model yourself. Everything is set up. All you have to do is go to “train your own” and follow the instructions. For more detailed information, you can learn more beforehand by reading our fine-tuning tutorial here. 
You can find even more tutorials in our blog-section.