StableTuner VS dreambooth-training-guide

Of all dreamboothing/finetuning implementations I tried I liked StableTuner the most. Might be worth giving it a shot to compare as well.

Largest I've gone is about 100 images for objects or people. I don't think it matters though, it can be a hassle setting up and resuming the training session each time if your doing small sessions. Stable Tuner can simplify all of this by helping you set everything up through their client installed locally. You can then easily do your training locally in short sessions or have it automatically packed up to be exported to colab or another gpu hosting service, also with the ability to train in short sessions. Its a smart way to manage large training projects like yours. It requires a bit of time setting up but most folks who have already played around with dreambooth should be able to navigate their way through easily enough. It has all the other training methods built into it too, including proper fine tuning https://github.com/devilismyfriend/StableTuner

Some people also like StableTuner: https://github.com/devilismyfriend/StableTuner

Stable Tuner seems like it's setup to allow training on regular/inpaint/depth models. https://github.com/devilismyfriend/StableTuner

StableTuner is an alternative to the sd_dreambooth plugin. It can do Dreambooth and Fine Tuning (I haven't tried this but I think it's embeddings) It uses diffusers but will convert between that and ckpt files, is for Windows/Nvidia, and uses a local app instead of a webapp. This is the only successful local Dreambooth I've done. You'll need to go to their discord for help but it's not hard to use.

Dreambooth is has better results in older commits. StableTuner is better for training : https://github.com/devilismyfriend/StableTuner

I dont have the time right now, but the rule of thumb for me was 80 unet learning steps for 1 image. Atleast 40 regularization images. Read more about regularization images here: https://github.com/nitrosocke/dreambooth-training-guide

This tutorial for dreambooth training has advice with regard to backgrounds which is probably also applicable to LORA. It recommends including images with solid, non-transparent backgrounds but not using them exclusively. Images that focus on the torso and face are probably most important unless your subject has very distinctive legs and feet. Removing other subjects is a must if you're training for a specific subject.

I found this, I'm going to go through this guide. Seems like I am using far too many images. https://github.com/nitrosocke/dreambooth-training-guide

Nitrosocke's guide already tells how much and what kind of images to use.

Dreambooth lets you do it in less than 100 images. https://github.com/nitrosocke/dreambooth-training-guide These folks say it's 5-15 to train on a person but I've not tested myself. https://www.reddit.com/r/StableDiffusion/comments/10tqy88/were\_launching\_a\_lightningfast\_dreambooth\_service/

See eg this tutorial for styles: https://github.com/nitrosocke/dreambooth-training-guide

The first time I tried to Dreambooth a style it went poorly. Then I found Nitrosocke's Dreambooth Training Guide and realized my problems were caused by a poorly redacted dataset.