My goal is to create a model using the Dreambooth extension in Stable Diffusion by training on a lot of images of the same person, leveraging CLIP interrogation to associate prompts with those images.
I used the CLIP interrogator (by pharmapsychotic) to batch process/generate prompts for 200+ images.
According to the docs, the colab notebook will either create "a desc.csv file in the same folder or rename the files to contain their prompts". In my case, all of the files were renamed from something like "789327838734.jpeg" to something like "a close up of a man with dreadlocks and a green hat, album art young thug, studio atmosphere, in a open-space working space,_1.jpeg". So it worked! All of my images have been renamed with prompts in the filename. (Though I did not see a desc.csv file saved in the same folder, but all of the files were renamed).
The docs for the colab notebook say "The renamed files work well for DreamBooth extension in the Stable Diffusion Web UI. "
I've installed the Dreambooth extension, but I just can't quite figure out how to use the new filename as part of the training...
It seems like I should be using [filewords] in the concepts input, but in this tutorial the use of [filewords] relies on .txt files associated with each image? I don't have a .txt file associated with each image. I just have a new file name.
How does the renamed file work with Dreambooth, exactly? I have not been able to find any information about this specific part of the step.
Before I start training on 200 images, which I imagine will be an investment of time and computing, I'm trying to find more information on how to use these new file names.
Many thanks
