It had a sort of surrealist look to it that I think threw a lot of people off. I don’t know how to explain it, but a lot of people thought it just looked off in one way or the other. For me, it was the lighting with the constant change in depth of field that made it look odd, but I still figured out it wasn’t AI after looking a bit closer.
Yes I work in the field and I’d say that overall this is somewhat accurate, denoising in particular has a direct impact on your pixels so at the core it’s “manipulated”, of course that doesn’t mean the content isn’t accurate but it’s definitely “processed”.
Then there’s all the heavier use of AI like remove tools, and partially using gen AI.
lattrommi@lemmy.ml
on 30 May 2025 01:03
nextcollapse
Do GIMP, Krita, Kdenlive or Inkscape use AI? I did not think they did, to the best of my knowledge. Maybe I’m missing something about AI assisted compression and correction, which I admit I’m not familiar with.
Does this only apply to digital media used in mainstream sources or does it mean everyone who uses editing software is using AI?
Powderhorn@beehaw.org
on 30 May 2025 07:32
nextcollapse
They all use algorithms – that’s what software is – but equating what’s been done for decades in software with AI is disingenuous. By this definition of AI, that was baked into Quark 3.3 and Photoshop 5 (not CS5, just 5).
What used to be done for decades, is being turned up to 100,000%. Instead of clever algorithms written directly by people, black-box AI algorithms and generative AI are being used to modify content so it fits better to the expectations of the old algorithms.
I wouldn’t be surprised if new compression algorithms came out in the next years, openly taking advantage of generative AI to recreate the “original image”… “original intent/concept?”
There are AI plugins for all of them… but they’re optional for now (2025). Kdenlive is working on integrating correction and background removal generative AI. Main offender is Adobe, which is the “standard” workflow for most media processing, and is forcing AI everywhere, including something as simple as color curves… then slapping a tag of “made using AI” in the output file. Inkscape is foremost a SVG editor, but Adobe Illustrator already has generative AI to allow stuff like rotating vector graphics “in 3D”, it’s only time for Inkscape to follow suit. Even Windows Notepad got some AI features recently 🤦
AI assisted compression and correction
JPG compression itself is a sort of “AI light”, where it analyzes chunks of an image for perceptual similarity, to drop “irrelevant” data. Adobe has added a feature to do that, but using AI in the analysis, tweaking/generating blocks so there are more similarities. It’s likely others will follow suit: “it’s lossy compression after all, right? …right?”
Lossy audio encoding (MP3, etc), also has a perceptual profile to increase block similarities, they’re adding AI there the same way as in images.
Videos… well, they’re a mix of images and audio, with temporal sequences already breaking images into key frames, intermediates, generated, etc. Generatively tweaking some of those to make them more similar, within perceptual limits, also improves compression.
Does this only apply to digital media used in mainstream sources or does it mean everyone who uses editing software is using AI?
Main issue lies at the source: cameras
Unless you’re using a large sensor professional camera, all the “prosumer” and smartphone sensors, are… let’s put it mildly… UTTER CRAP. They’re too small, with lenses too bad, unable to avoid CoC, diffraction, or chromatic aberration.
Before it even spits out a “RAW” image, it’s already been processed to hell and the way back. Modern consumer “better” cameras… use more AI to do a “better” processing job. What you see, is way past the point of whatever the camera has ever seen.
…and then, it goes into the software pipeline. ☠️
silentdon@beehaw.org
on 30 May 2025 19:43
collapse
I think you’re conflating “AI” with media processing. In most photo/video editing software that support it, you can use AI as a tool, but all it’s really doing is cutting down on the time it would take to do some tasks manually. That doesn’t mean it’s “AI” any more than it’s “AI” to crop a photo. Even film negatives need to be processed before anyone can see the photo.
I’m not saying AI good or bad, but I think it’s disingenuous to say that using AI to say, colour-correct an image or denoise a video, makes that image/video “AI”
I’m saying AI is being shoved into all steps of media processing.
Let me illustrate: this is an AI-focused, AI-corrected, AI-remastered, AI-lifted sticker of a photo of my cat… AI-cropped from a screenshot… that got AI-moderated the moment I uploaded it here.
threaded - newest
This is the greatest threat. It’s nof fake news, but the attack on truth itself?
Fake News does nothing but further dilute the trust in real news - the poison that numbs you to the knife in your back.
That threat hit about a decade ago. We’re completely lost in the sauce at this point
I don’t think Israel wants us to see Palestinians peacefully getting food either 😂
Why would anyone think it’s AI? The sound could well have been dubbed in though.
It had a sort of surrealist look to it that I think threw a lot of people off. I don’t know how to explain it, but a lot of people thought it just looked off in one way or the other. For me, it was the lighting with the constant change in depth of field that made it look odd, but I still figured out it wasn’t AI after looking a bit closer.
It was crappy stabilization and autofocus.
Nowadays, all digital media is becoming AI:
Doesn’t mean “generative AI”, but spotting the difference is only going to become harder and harder.
Yes I work in the field and I’d say that overall this is somewhat accurate, denoising in particular has a direct impact on your pixels so at the core it’s “manipulated”, of course that doesn’t mean the content isn’t accurate but it’s definitely “processed”.
Then there’s all the heavier use of AI like remove tools, and partially using gen AI.
Do GIMP, Krita, Kdenlive or Inkscape use AI? I did not think they did, to the best of my knowledge. Maybe I’m missing something about AI assisted compression and correction, which I admit I’m not familiar with.
Does this only apply to digital media used in mainstream sources or does it mean everyone who uses editing software is using AI?
They all use algorithms – that’s what software is – but equating what’s been done for decades in software with AI is disingenuous. By this definition of AI, that was baked into Quark 3.3 and Photoshop 5 (not CS5, just 5).
What used to be done for decades, is being turned up to 100,000%. Instead of clever algorithms written directly by people, black-box AI algorithms and generative AI are being used to modify content so it fits better to the expectations of the old algorithms.
I wouldn’t be surprised if new compression algorithms came out in the next years, openly taking advantage of generative AI to recreate the “original image”… “original intent/concept?”
There are AI plugins for all of them… but they’re optional for now (2025). Kdenlive is working on integrating correction and background removal generative AI. Main offender is Adobe, which is the “standard” workflow for most media processing, and is forcing AI everywhere, including something as simple as color curves… then slapping a tag of “made using AI” in the output file. Inkscape is foremost a SVG editor, but Adobe Illustrator already has generative AI to allow stuff like rotating vector graphics “in 3D”, it’s only time for Inkscape to follow suit. Even Windows Notepad got some AI features recently 🤦
JPG compression itself is a sort of “AI light”, where it analyzes chunks of an image for perceptual similarity, to drop “irrelevant” data. Adobe has added a feature to do that, but using AI in the analysis, tweaking/generating blocks so there are more similarities. It’s likely others will follow suit: “it’s lossy compression after all, right? …right?”
Lossy audio encoding (MP3, etc), also has a perceptual profile to increase block similarities, they’re adding AI there the same way as in images.
Videos… well, they’re a mix of images and audio, with temporal sequences already breaking images into key frames, intermediates, generated, etc. Generatively tweaking some of those to make them more similar, within perceptual limits, also improves compression.
Main issue lies at the source: cameras
Unless you’re using a large sensor professional camera, all the “prosumer” and smartphone sensors, are… let’s put it mildly… UTTER CRAP. They’re too small, with lenses too bad, unable to avoid CoC, diffraction, or chromatic aberration.
Before it even spits out a “RAW” image, it’s already been processed to hell and the way back. Modern consumer “better” cameras… use more AI to do a “better” processing job. What you see, is way past the point of whatever the camera has ever seen.
…and then, it goes into the software pipeline. ☠️
I think you’re conflating “AI” with media processing. In most photo/video editing software that support it, you can use AI as a tool, but all it’s really doing is cutting down on the time it would take to do some tasks manually. That doesn’t mean it’s “AI” any more than it’s “AI” to crop a photo. Even film negatives need to be processed before anyone can see the photo.
I’m not saying AI good or bad, but I think it’s disingenuous to say that using AI to say, colour-correct an image or denoise a video, makes that image/video “AI”
I’m saying AI is being shoved into all steps of media processing.
Let me illustrate: this is an AI-focused, AI-corrected, AI-remastered, AI-lifted sticker of a photo of my cat… AI-cropped from a screenshot… that got AI-moderated the moment I uploaded it here.
<img alt="" src="https://beehaw.org/pictrs/image/cd553cd7-c350-43d5-ba4e-f20faa431575.jpeg">
Also please use google maps:
maps.app.goo.gl/GY88ymDx6TkiZ1Dx9
That’s a school made into gravel.