Hello, this is author of Illustrious series.

#2
by AngelBottomless - opened

And I want to clarify we have definite open source plan, including official uploads - and this is not the way to make it sustainable.

Sorry for making everyone confused - we acknowledged that everyone have been wondering what is happening currently. We will make announcements very soon.

We will showcase models sequentially, also making time for everyone to test, adapt and find out compatibility, and also preparing inference setups and so on - including open sourcing models in progressive manner.

We had along time of sleep - this is being late, but however, as you can see - there are too many stuffs to be explored. We want this interesting research & release sustainable, so please do not make it stop.

One thing is definitive, unless similar situations are repeated - you'll see official ones soon.

As everyone knows, we had similar incidents with our "v0.1 beta test" - we clarified that the weights will be released officially, however which was distracted by similar incidents, which made us sleep forcibly for several months.

Please do not make it happen again.

If you wonder about "payments", we have to discuss more - but model training wasn't free, yes, it spent a lot of money - but still, we promise we will take similar approaches for development & showcase & release, even extended to next-generation models and so on.

Downloaded, cry about it

Dear AngelBottomless,

Thank you for your response, and I genuinely appreciate the effort and vision behind your work. The unpredictability of these events is understandable, and while it disrupts expectations, it also presents an opportunity to reassess and strengthen your future release strategy.

Given the high-stakes nature of upcoming models, I encourage you to treat this as a valuable stress test—one that can help refine a more resilient approach to distribution, sustainability, and open-source integration. If history has shown us anything, it's that unexpected events often force necessary evolution. The key is ensuring that disruptions don’t stall progress but instead shape a more robust, adaptive framework for the next phase of releases.

Looking forward to seeing your future models and how they shape the landscape.

Best,
Liberata Team

Ledditors pretending they care about using artists work for free is the best part of this situation lol

model training wasn't free

yet you took the training materials for free without permission

Downloaded, cry about it (2)

Don't want it not even for free, shitty sjw model with loli/lolicon content filtered out, might as well bankrupt

roviding the 1.0 model for paid download is a misstep. You should emulate novel.ai and offer generation services of the 1.0 model on your own website. Civitai.com won't pay you; their user base tends to seek freebies, which are not your target customers. It's better to straightforwardly explain that you're a profit-oriented business rather than offering more open-source explanations. Open-source contributions often do not yield returns.
I respect commercial endeavors; it makes sense to receive reasonable compensation for one's labor. It is improper to use morality to coerce others into open-sourcing their work.

@AngelBottomless your message says literally nothing, stop writing corporate bullshit.

You've made a mistake - take responsibility and apologize, that's all you can do at this point. You and your company have just started the process of slow destruction of opensource imagen, and you (including you personally) will always be guilty of that. Live with it.

The way I see this happening, I'd say that it was bound to happen. Not even larger AI companies are immune to weights getting redistributed/leaked, Miqu 70B being a prime example, and it is close to impossible to prevent that. The arxiv white paper described the Illustrious models as "an Open Advanced Illustration Model" and releasing as a paid model, even for early access, subverts this expectation. Paywalling early releases of open models or open-source software is already known to cause incidents, the (now defunct) yuzu emulator and Pineapple case being one of them. I was quite skeptical and suspicious about the release plan described in that paper and it turned out like this...
Despite that, people that are willing to support you will pay for the model, even though the weights are in the wild.
Please take this incident as a global reflection on monetization of open software and models. While training models has a cost, I believe this way of recovering costs is less than ideal.

"which made us sleep forcibly for several months" What is that even supposed to mean?

And I want to clarify we have definite open source plan, including official uploads - and this is not the way to make it sustainable.

Sorry for making everyone confused - we acknowledged that everyone have been wondering what is happening currently. We will make announcements very soon.

We will showcase models sequentially, also making time for everyone to test, adapt and find out compatibility, and also preparing inference setups and so on - including open sourcing models in progressive manner.

We had along time of sleep - this is being late, but however, as you can see - there are too many stuffs to be explored. We want this interesting research & release sustainable, so please do not make it stop.

One thing is definitive, unless similar situations are repeated - you'll see official ones soon.

As everyone knows, we had similar incidents with our "v0.1 beta test" - we clarified that the weights will be released officially, however which was distracted by similar incidents, which made us sleep forcibly for several months.

Please do not make it happen again.

If you wonder about "payments", we have to discuss more - but model training wasn't free, yes, it spent a lot of money - but still, we promise we will take similar approaches for development & showcase & release, even extended to next-generation models and so on.

So my question is:
Does this make 0.1 and 1.0 both more illegal than NAI leaks?
Same with NoobAi ..
Like why don't you go after all the sites that have it.
TensorArt is .. deplorable no offense, their owner stalked me and refused to take down NSFW child content, where at least 10 other asian sites WILL attempt on all bounds to find it and get rid of it when it's found.
Civitai has this model, i'm sure Shakker will i'm sure EVERYONE ELSE WILL.

And a 10k buzz limit on a STOLEN model? :3
Y'all realize you will not make much on Civitai if anything at all - cannot use that buzz, cannot get revenue -- because tehe CC program is closed dick move really XD

"which made us sleep forcibly for several months" What is that even supposed to mean?

They're korean so it's probably an expression in that language. The entire response seems translated too

Yea look i'm not doubting their need to make back money i just find it funny that they're not doing like
... y'know a website with merch, patreon etc--
Becuse honestly the minute you put it on a paid site someoen's gonna make the call to download it and put it up if your liscence. denotes it

Sign up or log in to comment