Some of those demos are pretty engaging, I won't lie. As just part of the creative flow, and not the complete e2e process, this will, I suspect, be impossible to detect, as human curation will...
Some of those demos are pretty engaging, I won't lie. As just part of the creative flow, and not the complete e2e process, this will, I suspect, be impossible to detect, as human curation will eliminate all the usual edge cases like poor hands.
I imagine tooling like this will soon become present in all complex photo or video editing tooling. Since things like StableDiffusion are open source, even Gimp will, if it doesn't already (it probably does), have plugins for this.
The current Firefly generative AI model is trained on a dataset of Adobe Stock, along with openly licensed work and public domain content where copyright has expired.
On one hand, it addresses common criticisms around the training of AI image generators, but on the other hand, if this becomes the standard that people expect, then it may become that the only...
On one hand, it addresses common criticisms around the training of AI image generators, but on the other hand, if this becomes the standard that people expect, then it may become that the only socially acceptable AI image generators are the ones controlled by corporations with big pockets, and image generators will be unable to reference so many things from popular culture that people will want to use them for.
My guess/hope about the way this ends up working out is that the first couple large popular commercial products to integrate image generators play it safe like this, causing controversies around image generator training sets to die down despite popular open source image generators like Stable Diffusion changing little about how they work, people get comfortable with the socially-acceptable corporate tools, and then later come around to appreciating the open qualities of Stable Diffusion.
YouTube is a messy compromise between copyright holders and the free sharing of video. Maybe something else like that will be built? Adobe claims to be coming up with a way to pay contributors to...
YouTube is a messy compromise between copyright holders and the free sharing of video. Maybe something else like that will be built?
Adobe claims to be coming up with a way to pay contributors to Adobe Stock.
From the announcement they said that they would be making it clear when something had been generated with AI, and letting people opt out of their content being included in the training by applying...
From the announcement they said that they would be making it clear when something had been generated with AI, and letting people opt out of their content being included in the training by applying a tag.
I also recall hearing about AI ethics boards and stuff that Adobe is participating in, although I don't remember the details. It may just be lip service, but it at least seems like Adobe is trying to do well with it.
I got to play with this in the internal beta before it was announced. Went in not really caring much about it since its another generative AI tool, but found myself having a lot of fun seeing what...
I got to play with this in the internal beta before it was announced. Went in not really caring much about it since its another generative AI tool, but found myself having a lot of fun seeing what the other 30k people working at Adobe did with it. It has the usual issues with generative AIs (freaky hands, weird conceptions of things, etc.) but seems to really excel at generating content you'd use for illustrations and adding effects to existing things. The impression it left on me was that it would be a very cool tool for existing illustrators and designers to use to enable their creativity and streamline their workflow, and not as much a general AI made to handle every problem imperfectly.
Also glad to see the response here so far isn't negative, unlike for most things Adobe does 👀
Some of those demos are pretty engaging, I won't lie. As just part of the creative flow, and not the complete e2e process, this will, I suspect, be impossible to detect, as human curation will eliminate all the usual edge cases like poor hands.
I imagine tooling like this will soon become present in all complex photo or video editing tooling. Since things like StableDiffusion are open source, even Gimp will, if it doesn't already (it probably does), have plugins for this.
Looks like they are avoiding the legal issues:
On one hand, it addresses common criticisms around the training of AI image generators, but on the other hand, if this becomes the standard that people expect, then it may become that the only socially acceptable AI image generators are the ones controlled by corporations with big pockets, and image generators will be unable to reference so many things from popular culture that people will want to use them for.
My guess/hope about the way this ends up working out is that the first couple large popular commercial products to integrate image generators play it safe like this, causing controversies around image generator training sets to die down despite popular open source image generators like Stable Diffusion changing little about how they work, people get comfortable with the socially-acceptable corporate tools, and then later come around to appreciating the open qualities of Stable Diffusion.
YouTube is a messy compromise between copyright holders and the free sharing of video. Maybe something else like that will be built?
Adobe claims to be coming up with a way to pay contributors to Adobe Stock.
From the announcement they said that they would be making it clear when something had been generated with AI, and letting people opt out of their content being included in the training by applying a tag.
I also recall hearing about AI ethics boards and stuff that Adobe is participating in, although I don't remember the details. It may just be lip service, but it at least seems like Adobe is trying to do well with it.
I got to play with this in the internal beta before it was announced. Went in not really caring much about it since its another generative AI tool, but found myself having a lot of fun seeing what the other 30k people working at Adobe did with it. It has the usual issues with generative AIs (freaky hands, weird conceptions of things, etc.) but seems to really excel at generating content you'd use for illustrations and adding effects to existing things. The impression it left on me was that it would be a very cool tool for existing illustrators and designers to use to enable their creativity and streamline their workflow, and not as much a general AI made to handle every problem imperfectly.
Also glad to see the response here so far isn't negative, unlike for most things Adobe does 👀