Humans

To generate avatars or photorealistic results. Remember to caption the image correctly when uploading the pictures.

✅ Do’s

At least 15 photos

Majority (at least 10) closeup photos of face (various styles, expressions, lighting, poses, etc.)

Various lighting, poses, cameras, clothes

❌ Dont’s

No more than 10% of photos in black and white

No more than 10% of photos taken from a bad camera (mobile)

No more than 20% of photos with the accessories (glasses, hat, etc.)

❌ Not too far away in the photo

❌ No photos in mobile format 9/16

❌ No 2 photos very similar or identical

❌ No photos with other people

Objects

Either to generate variations of an object or to generate the object accurately in different places, backgrounds. Note that AI is currently not good at positioning objects into space with precision as spatialization is one of the hardest things with our current models

✅ Do’s

• At least 5 photos
• Variation or angles - no 2 similar photos.
• If looking for consistency,
• If looking for variability, try having different environments