jusjakn said: Love it! Top one has to be AI though. Way too crisp. Your wife is so pretty irrespective.
I certainly hope they are NOT AI, as it's good to see such a nice (skirted ) woman getting messy.
There are quite a few tools that can detect ai images or deep fakes. These detect the messy images here as fake. Also, skirtpie posted the same thread on the ai group https://umd.net/groups/old-pics-4
I think a lot of them are image-to-image manipulations. They start with a real photo and add a mess. And yes, you can take a photo of any real person and add a mess. Upscaled photos are also detected as fake.
There have been a few people on umd posting a bunch of images and passing them off as real when several tools detect them as fake.
jusjakn said: Love it! Top one has to be AI though. Way too crisp. Your wife is so pretty irrespective.
I certainly hope they are NOT AI, as it's good to see such a nice (skirted ) woman getting messy.
There are quite a few tools that can detect ai images or deep fakes. These detect the messy images here as fake. Also, skirtpie posted the same thread on the ai group https://umd.net/groups/old-pics-4
I think a lot of them are image-to-image manipulations. They start with a real photo and add a mess. And yes, you can take a photo of any real person and add a mess. Upscaled photos are also detected as fake.
There have been a few people on umd posting a bunch of images and passing them off as real when several tools detect them as fake.
So, should someone flag as synthetic? Still figuring out the rules myself, but seems like passing off AI images as real should be a no-no.
jusjakn said: Love it! Top one has to be AI though. Way too crisp. Your wife is so pretty irrespective.
I certainly hope they are NOT AI, as it's good to see such a nice (skirted ) woman getting messy.
There are quite a few tools that can detect ai images or deep fakes. These detect the messy images here as fake. Also, skirtpie posted the same thread on the ai group https://umd.net/groups/old-pics-4
I think a lot of them are image-to-image manipulations. They start with a real photo and add a mess. And yes, you can take a photo of any real person and add a mess. Upscaled photos are also detected as fake.
There have been a few people on umd posting a bunch of images and passing them off as real when several tools detect them as fake.
So, should someone flag as synthetic? Still figuring out the rules myself, but seems like passing off AI images as real should be a no-no.
I did report a few. One user used to post many synthetic images without the tag, but now does. A lot of people commented on the images seemed to think they were real.
skirtpie's images may be upscaled which is allowed to be posted without the synthetic tag according to the rule:
Content that's had its quality enhanced or upscaled by AI does not need to be tagged synthetic because the scenario and models are still real, not made-up.
Some of the photos are 99% likely to be ai according to some tools, and it even says it is 96% likely to be generated by GPT-4o.
Another user did not use the synthetic tag, claimed an image was real, it was reported, marked synthetic, and she claimed that she replaced her face to conceal her identity. The tool said it was 99% likely ai and 98% likely Stable Diffusion.
Coming back to skirtpie's images. Clearly some are old photos and some are larger, sharper, and more detailed ones. The tool says those larger ones are 99% likely ai, and it is 93% likely MidJourney. The lower resolution ones are 98% likely real. The tool I use is https://sightengine.com/detect-ai-generated-images
Also, there are other processes other than upscaling that adds nuance to how content is labeled. Outcropping is a process that adds more background. The subject could be original but ai is used to replace the background. The background may be original but ai is used to replace or alter the subject. It could be modified to give the person sunglasses to conceal their identity. It could be used to replace something in the background. There is sort of a Ship-of-Theasus paradox to it.
jusjakn said: Love it! Top one has to be AI though. Way too crisp. Your wife is so pretty irrespective.
I certainly hope they are NOT AI, as it's good to see such a nice (skirted ) woman getting messy.
There are quite a few tools that can detect ai images or deep fakes. These detect the messy images here as fake. Also, skirtpie posted the same thread on the ai group https://umd.net/groups/old-pics-4
I think a lot of them are image-to-image manipulations. They start with a real photo and add a mess. And yes, you can take a photo of any real person and add a mess. Upscaled photos are also detected as fake.
There have been a few people on umd posting a bunch of images and passing them off as real when several tools detect them as fake.
So, should someone flag as synthetic? Still figuring out the rules myself, but seems like passing off AI images as real should be a no-no.
I did report a few. One user used to post many synthetic images without the tag, but now does. A lot of people commented on the images seemed to think they were real.
skirtpie's images may be upscaled which is allowed to be posted without the synthetic tag according to the rule:
Content that's had its quality enhanced or upscaled by AI does not need to be tagged synthetic because the scenario and models are still real, not made-up.
Some of the photos are 99% likely to be ai according to some tools, and it even says it is 96% likely to be generated by GPT-4o.
Another user did not use the synthetic tag, claimed an image was real, it was reported, marked synthetic, and she claimed that she replaced her face to conceal her identity. The tool said it was 99% likely ai and 98% likely Stable Diffusion.
Coming back to skirtpie's images. Clearly some are old photos and some are larger, sharper, and more detailed ones. The tool says those larger ones are 99% likely ai, and it is 93% likely MidJourney. The lower resolution ones are 98% likely real. The tool I use is https://sightengine.com/detect-ai-generated-images
Also, there are other processes other than upscaling that adds nuance to how content is labeled. Outcropping is a process that adds more background. The subject could be original but ai is used to replace the background. The background may be original but ai is used to replace or alter the subject. It could be modified to give the person sunglasses to conceal their identity. It could be used to replace something in the background. There is sort of a Ship-of-Theasus paradox to it.
Thanks for the explanation! Much appreciated. There can be lots of gray to it. I wasn't sure to what degree originally messh images were upscaled or enhanced vs someone taking a picture of a lady and adding a pie to the face. There wasn't a lot of context to go on, so I figured I'd ask.