41–50 of 50 posts
What are people using for self hosting these? I tried ollama with open-webui and it didn't support image generation at all.
Earlier quoted context omitted.
What is Krea's approach to content such as pornography and gore? It's been frustrating to see all of the leading models take a very hard line on excluding vice content, even when it is perfectly legal, in the name of safety.
> What is Krea's approach to content such as pornography and gore? You mean whether it's censored? Of course it is. They even say so themselves[1]: > The open source version needed to go through some alignment training so there might be some inconsistencies between closed / open version. The "alignment training" is pretty much a code word for censorship/lobotomy, because unlike the API version they can't slap on a sa…
I disagree.
Grok is a hosted service that can use classifiers and has been personally guided by Elon Musk directly before, that was choosing to generate and distribute child porn for users. Then after getting in trouble, had the gall to continue doing it behind a paywall.
Results are in! This is a really impressive showing especially given how fast the Turbo model at 8 steps. The only locally hostable model that managed to outperform it was Ideogram 4 which is significantly slower (think minutes vs seconds). It did fall to the usual “model killers”: the nine-pointed star, Count Rugen, the overcrowded flat Earth . But overall, it really punched above its weight class, scoring the highe…
I'd never heard of text to image model killers so I had a good chuckle at this. Such oddly specific things for us to arrive at as a test method
> 2.3 Revenue Threshold for Commercial Use. Commercial Use under this Agreement of the Krea Model, Derivatives, or Outputs is permitted only if you (including all affiliated entities under common ownership or control) have total company-wide annual revenue of less than one million United States dollars ($1,000,000 USD), calculated on a trailing twelve-month basis and including all revenue from all sources. If you meet or exceed this threshold, you must obtain a separate enterprise license from Krea prior to any Commercial Use. If your revenue meets or exceeds this threshold at any time during your use of the Krea Model under this Agreement, you must immediately cease Commercial Use and contact Krea. Enterprise license inquiries may be directed to opensource@krea.ai.
> 4.1 General Restrictions. You shall not, and shall not permit any third party to: (a) Use the Krea Model, any Derivative, or any Output in violation of applicable law, regulation, this Agreement, or the Acceptable Use Policy;
> 4.2 Content Filtering Requirement. You must implement reasonable and appropriate Content Filter measures to detect, prevent, and mitigate the generation or distribution of prohibited, harmful, or unlawful content through your deployment of the Krea Model or any Derivative. Such measures may include, but are not limited to: (a) open-source content classifiers, such as Falconsai/nsfw_image_detection, NudeNet, or CompVis safety checker; (b) commercial content moderation APIs, such as Hive Moderation or Microsoft Azure AI Content Safety; (c) manual human review processes; and/or (d) any combination of the foregoing or other technically appropriate measures.
> 4.4 Acceptable Use Policy Compliance. You must comply with the Acceptable Use Policy, which is incorporated herein by reference.
The acceptable use policy is on the website (https://www.krea.ai/krea-2-use-policy) and includes:
> You shall not use or allow others to use the Krea 2 Raw Model or Krea 2 Turbo Model, any Derivative, or any Output for any of the following purposes:
> (8) Circumventing or removing any safety measures, usage restrictions, content filters, content provenance, or watermarking mechanisms implemented by Krea or any deployer;
What are people using for self hosting these? I tried ollama with open-webui and it didn't support image generation at all.
Earlier quoted context omitted.
> What is Krea's approach to content such as pornography and gore? You mean whether it's censored? Of course it is. They even say so themselves[1]: > The open source version needed to go through some alignment training so there might be some inconsistencies between closed / open version. The "alignment training" is pretty much a code word for censorship/lobotomy, because unlike the API version they can't slap on a sa…
> If they don't do this then the next post we'd see is them being in hot water. (Remember what happened with Grok? That's what happens when you don't take a hard line here.) I disagree. Grok is a hosted service that can use classifiers and has been personally guided by Elon Musk directly before, that was choosing to generate and distribute child porn for users. Then after getting in trouble, had the gall to continue…
This is a slant, a lie of presumption and ignorance.
Ignorance of what "open" means, presumption of assigning intent without validation.
Some believe that lobotomizing models leads to suboptimal results overall. Some also believe that the user, not the tool is responsible.
In these contexts, you'd better call a paint manufacturing company "choosing to enable child porn".
I am so sick of US style politics, with its insane labelling by both parties, and their worshippers, with intent to besmirch on every breath, because someone doesn't agree with them in some narrow political context.
From outside the US, most can barely tell the difference between a democrat and a republican. Both sides are slimy, smarmy, dishonest, without honour, and spew lies and bull about anyone they don't agree with.
It's absolutely disgusting.
Results are in! This is a really impressive showing especially given how fast the Turbo model at 8 steps. The only locally hostable model that managed to outperform it was Ideogram 4 which is significantly slower (think minutes vs seconds). It did fall to the usual “model killers”: the nine-pointed star, Count Rugen, the overcrowded flat Earth . But overall, it really punched above its weight class, scoring the highe…
> It did fall to the usual “model killers”: the nine-pointed star, Count Rugen, the overcrowded flat Earth. I'd never heard of text to image model killers so I had a good chuckle at this. Such oddly specific things for us to arrive at as a test method
What’s more surprising to me is that, unlike the “pelican riding a bicycle” whose objectivity has been slightly compromised as newer models have incorporated it into their training data, the arbitrary-point star has been wiping models out ever since the early days of Flux back in 2024.
I personally love the test because it's something that even an elementary school child with no artistic experience at all can do, but state of the art models struggle heavily.
What are people using for self hosting these? I tried ollama with open-webui and it didn't support image generation at all.
* https://github.com/LostRuins/koboldcpp/releases/tag/rolling
Model weights:
* https://github.com/leejet/stable-diffusion.cpp/blob/master/d...
In the launcher GUI, select 'Image Gen' and set Krea2 Turbo gguf as the image model, Qwen3 VL 4b gguf as the T5 model, and Wan2.1 safetensors as the VAE.
Earlier quoted context omitted.
> If they don't do this then the next post we'd see is them being in hot water. (Remember what happened with Grok? That's what happens when you don't take a hard line here.) I disagree. Grok is a hosted service that can use classifiers and has been personally guided by Elon Musk directly before, that was choosing to generate and distribute child porn for users. Then after getting in trouble, had the gall to continue…
that was choosing to generate and distribute child porn for users. This is a slant, a lie of presumption and ignorance. Ignorance of what "open" means, presumption of assigning intent without validation. Some believe that lobotomizing models leads to suboptimal results overall. Some also believe that the user, not the tool is responsible. In these contexts, you'd better call a paint manufacturing company "choosing to…
Furthermore, even if I don't make the painting, if I see it and distribute it anyway, I'm still distributing child porn. You don't get to just say "whoopsies, I didn't make it, how could I possibly know the open and visible image was illegal?"
Grok isn't making paint. It's not making monitors. It doesn't send you graphics software. It itself produced pornographic images of real children and distributed them to pedophiles for profit.
And even if we wanted to give them the benefit of the doubt, after this came to light and they got in trouble, they disabled this capability for free users while allowing paid users to continue. They knew what they were doing at that point, so no more ignorance plea.
At least open model producers are just distributing software that can produce illicit images. But because they don't actually do it, they can wash their hands of responsibility.
Results are in! This is a really impressive showing especially given how fast the Turbo model at 8 steps. The only locally hostable model that managed to outperform it was Ideogram 4 which is significantly slower (think minutes vs seconds). It did fall to the usual “model killers”: the nine-pointed star, Count Rugen, the overcrowded flat Earth . But overall, it really punched above its weight class, scoring the highe…