Spaces:
Running
Public domain
Is there a way to have the bot search exclusively in images which are unambiguously public domain? I'm assuming it's using a Google image API, so is there a command you can use in the prompt to change search settings to public domain?
That's a major issue with most ai text to image generators. The images they are trained with come from the internet, but are often not public domain. Which means that what you get is derivative of existing artists and designers. It's not searching, it's creating a new image from existing associations. But more often than not it's not entirely new. You would have to sort out billions of images and retrain it, which is not something that happens each time a user requests output. If you do a reverse image search of the results, you can sometimes find the source material.
That's kind of my point but if they're using a script to get this data from, for example, Google Image Search, it's possible to filter results to only search images tagged under the Creative Commons licensure agreement.
仙境般的风景，拟人化了一个超可爱的白色仙女蛇宝宝，拿着玉如意，飞来飞去，穿着优雅的斗篷，露出甜美的笑容，白色的皮毛，明亮的大眼睛，毛茸茸的尾巴，云朵轻轻漂浮，极其细致的3D动画，超逼真，超细致，奢华，电影灯光，超清晰的细节，超清晰的材质，特写，复杂的纹理，辛烷值渲染，Zbrush，8K，VRAY超逼真的14247 --v第四章第二节第三节
That's kind of my point but if they're using a script to get this data from, for example, Google Image Search, it's possible to filter results to only search images tagged under the Creative Commons licensure agreement.
For sure. Any data set can be used. But I'm not sure that the results from Craiyon aren't public domain, since they have been often been significantly altered from their original form.
AI itself can not copyright anything it produces because it is not a human. It is just a tool, so the output can be copyrighted by the user so far he took some control about the image generation process.