Train models

Expert mode

ResultsSee results

Base model

?What type of results will you want to generate with your model?
Leave "Default" if you aren't sure.

Training subject

?What are you training this model on?
Indicate if you are using pictures of a person, an animal, ...

Model type

?Stable Diffusion 1.5 or Stable Diffusion XL

Base model

?Select the base model to finetune.
It will be trained on the images you upload.

Training resolution

?Resolution at which the model is trained.
SDXL can generate 1024x1024 images even when trained at 512x512.
512x512 will give better results if the original training data has limited resolution.

Training steps

?Number of steps to finetune the model for.
Between 100 and 500,000.
Each 5,000 training steps costs 30 tokens.

Learning rate

?How aggressively the model will be trained.
1e-5 is a good default.
Between 1e-7 and 1e-4.

Instance prompt

?The prompt used to finetune on the images you upload.
"photo of ukj person" is a good default for people.

Image cropping

?Images are cropped to 1024x1024 before finetuning SDXL.
Face Crop attempts to find a face and uses Center Crop if it can't find any.

Checkpoints

?You can extract a LoRA from the trained model, at no extra cost.
LoRA files are in Compvis format (AUTOMATIC1111-compatible).

Checkpoint format

?File format of the finetuned checkpoint.
CompVis contains a single checkpoint file.
Diffusers contains separate files for the UNet, VAE, tokenizer etc in a tarball.

Callback URL

?URL that will be called when the job finishes.

Offset Noise

?Not recommended if you don't specifically need this.
Helps training models that can generate very dark and very bright images.

Image captions

?Use image captions instead of the instance prompt.
Expects a JSON file, check the documentation for details.

Upload Captions

Sample prompt for your model: "photo of ukj person"

Upload 5 to 15 training images.

Latest jobs

	Job	Base model	Steps	Images	LR	Instance prompt	State	Duration	Download
Authenticating...

See all training jobs

dreamlook.ai

Social Media

Discord
LinkedIn
Twitter
Affiliate

Documentation

Pricing
API integration
AUTOMATIC1111
DiffusionBee

Legal

Privacy Policy
Terms of Use
Data Processors

We use Cookies

By using this website, you agree to our use of cookies. We use cookies to allow our website to run effectively.