A thumbnail is judged at postage stamp size
Three words maximum, one face, one strong colour. Anything more disappears in the feed.
Text is why the model matters
Most image models mangle letters. This one renders type cleanly, which is the entire job here.
Four layouts that keep working
Face and text, before and after, a question, or a big number for a list video. Pick one and change only the words.
FAQ
What size is the output?
Generate and upscale. YouTube wants 1280 by 720, so start larger and downscale.
Can I use my own face?
Yes, upload it. The layout builds around your photo.
Will the text be spelled right?
Usually. Keep it to three words and check before uploading.