Create your AI model
To generate images with your face the AI needs to learn from your photos and getting a good set of images is the most important part! Select the best ones and upload them bellow
What images should I use?
Bad Image Set
- BAD FACE VISIBILLITY
- BACKGROUD AND OUTFIT REPETITION
- POOR VARIABILITY (POSES, COLOR, LIGHT)
- LOW RESOLUTION
- SAME FACIAL EXPRESSIONS
Good Image Set
- GOOD FACE VISIBILLITY
- VARIETY IN BACKGROUD AND OUTFIT
- GOOD VARIABILITY (POSES, COLOR, LIGHT)
- HIGH RESOLUTION
- DIVERSE FACIAL EXPRESSIONS
50-60% Close Face-Shoulders
40-30% Upper Body
10-20% Fullbody
50-60% Close Face
40-30% Upper Body
10-20% Fullbody
Crop & Upscale
Select the best parts:
Crop the images to a 1:1 aspect ratio, focusing on the subject and mantaining the proportions shown above. The results will be similar to the images uploaded
Links to Crop Images:
Important tips:
The ONLY thing in common between your images shoud be the concept you want to train on the AI (the subject’s face)
- Quality is more important than quantity
- Choose images with diferent background, clothing, light, pose or any other thing you don’t want the AI to “always” reproduce
- The images should be as close as possible to the kind of images you’re trying to make
Upscale (optional):
If the image quality is too low you can try upscaling using AI before cropping to 512×512. Better images = better results
Crop Images:
First crop the images to a 1:1 aspect ratio (perfect square), focusing on the subject and mantaining the proportions shown above. Keep in mind that the results will be similar to the images uploaded
The images should be 512×512 pixels. DO IT HERE
Select the best images
Before uploading the images in the section below, a pop-up window will open to crop the images to a 1:1 aspect ratio
The ONLY thing in common between your images shoud be the concept you want to train on the AI (the subject’s face) so follow this tips:
- Most of your images should be a head and shoulders or torso frame, with a few close-ups of the face and a couple of full body shots if you wish as well.
- Quality is more important than quantity and you can get very good results with 10 to 30 images.
- Choose images with diferent background, clothing, light, pose, facial expressions or any other thing you don’t want the AI to “always” reproduce.
- The images should be as close as possible to the kind of images you’re trying to make.
- Modify with the proportions shown above (head/torso/fullbody) according to the type of images you want to generate.
Bad Images? No problem
If the images are of poor quality or with little variability, there are several methods to improve them.
Upscale and Enhance (optional):
When you purchase (2) “credits” you are also awarded with (1) “tool credits” (you can purchase them separately) which you can also use to upscale and enhance your images before uploading them if the quality is too bad.
If the image quality is too low you can try upscaling using AI before uploading them. Better images = better results
Upload and wait!
Upload your images, get your models and examples in 2 to 3 hours and start generating here or on your PC
Upload 3-100 images of someone mantaining the desired or recommended proportions shown above.
Very good results with 20-25 images. Quality is more important than quantity
Before uploading the images here a pop-up window will open to crop the images, keep in mind that the results will be similar to the cropped images so make sure to follow the tips shown above
- *DON’T MODIFY THE TOKEN unless you know how to set a good training token for a Dreambooth based training (single word)
- The optimal training class can vary depending on the subject and training images, the results could be better with “person” or “man/woman” training class
1 Custom Model = 25 credits
You have:
0 credits
