To add to wtf’s comment, ai tends to embed a watermark so that when you train a model, you know if it was trained using legit photos or ai generated photos.
That’s not to say you can’t disable the watermark like others have disabled the NSFW filter, but you’d have to fire up your own instance or find a shared one with those features stripped.
Like wtf said, it’s more of a guideline. If I had to guess, the human factor went up since the photo has the twitters handle overlayed so any manipulation to the photo will sometimes mess with the accuracy. Just things you need to know when using such a tool.
To add to wtf’s comment, ai tends to embed a watermark so that when you train a model, you know if it was trained using legit photos or ai generated photos.
That’s not to say you can’t disable the watermark like others have disabled the NSFW filter, but you’d have to fire up your own instance or find a shared one with those features stripped.
Like wtf said, it’s more of a guideline. If I had to guess, the human factor went up since the photo has the twitters handle overlayed so any manipulation to the photo will sometimes mess with the accuracy. Just things you need to know when using such a tool.
https://isitai.com/how-it-works/