tal

This is a remake of an image that I submitted about a year ago here, this time higher-resolution, in 16:9 aspect ratio, photographic and done with the Flux-derived Newreality model.

To repeat my comment then:

Some years back, I gathered up all the scary images I could find, and looked for commonalities, tried to figure out what was “frightening” to people. One pretty consistent element was a wide, toothed smile.

While I think that the earlier image is tough competition for this one in terms of visual impact, when I created the earlier image, I was much more lax in terms of what I was willing to accept from the generator -- I just wanted it to look scary, and to have a wide, toothed mouth. That is, a lot of getting the image was a dice roll; I had to generate a bunch of images, take the best.

But this time around, I have enough control from the model's ability to take natural-language description to be able to intentionally try to specifically target something like the original image, which is a stupendous improvement. Was just like an arrow to a very specific image that I wanted, just took a few minutes and iterations. Fantastic.

I hadn't gotten SD Ultimate Upscaler (using SwinIR_4x) set up with a Flux-based model until this image, so I'd been just putting out images at a quarter this resolution -- my monitor is 2560x1440, so this is a fullscreen image for me -- until now.

ComfyUI lets you "disable" a node in a workflow -- it's an option available in the right-click context menu when clicking on a node -- so if you use the posted workflow, you can just "disable" the upscaler part of the workflow until you actually care about its output, then re-enable it, as it's time-consuming.

I hopefully now have my general vanilla workflow to use for most images back up and running, which I haven't really fully since moving off Automatic1111. :-)

"The 'narco thugs' have no limits any more... These shootouts aren't happening in South America, they're happening in Rennes, in Poitiers... we're at a tipping point," Retailleau said Friday.

I mean, they're happening in South America too.

Just that control of distribution turf in France is also something worth money and thus worth fighting over, just like control of smuggling pipelines out of Latin America. And because it's all illegal activity, it's not like competing distributors can go to law enforcement or the courts to mediate things, so it goes back to the law of the jungle whenever conflict resolution has to happen.

I'm assuming that they're Bluetooth, as if they were wired, the problem couldn't really come up.

If they're still paired to a device, crank the volume up and try playing something loud enough that you can hear it.

For Bluetooth devices that are powered on and responding to queries for nearby Bluetooth devices, you can also try asking a device that can pair with Bluetooth devices and show signal strength, like a laptop, to query for nearby devices, can kind of use to "home in" on the device.

I rarely use speakers, as my environment is one where I might bother others easily. On the rare occasions that I do, I use some small old USB Logitech speakers from ages back. I have a surround sound speaker setup with subwoofer and such sitting around, but don't bother to plug it in, since I'm just not going to use it.

I mostly use headphones, though I've got an elaborate mass of mixers, sound interfaces, headphones, and multiple clock-synchronized powered mics hooked up to the computer.

I will say that small speakers have come a hell of a long way since the 1990s, when I remember them consistently sounding kind of tinny. They pretty much all sound great to me today.

I don't have a Google account precisely because I want to make it harder for Google to build a profile on me, so I probably don't see profile-based data other than the limited extent to which Google uses IP geolocation as an input to results.

which means only one person can win and any vote for someone who isn’t in the top two is pretty much a waste.

Parliamentary systems basically do the same thing as presidential systems, just in a different way.

There are only two really viable parties, and other parties can only really influence things via the spoiler effect in the US.

Yes. However, there are also very few actually viable party coalitions in most parliamentary systems. Like, the far-left party probably isn't going to enter into coalition with the far-right party. And neither has anywhere near enough support to actually determine the executive. Any coalition they enter into is going to mandate a lot of compromises from what their particular party program -- what we in the US typically call a "party platform" is. So...they aren't really an option for running the executive, even if they show up on the ballot.

In a parliamentary system, the parties make their promises to the public. Then the vote happens. Then there's some horse-trading, and parties throw out some -- not known to the public at the time of election -- of their electoral promises, and create a coalition.

It's true that in the US system, you basically only have two viable parties...but that's because in the US, parties are more analogous to party coalitions in some parliamentary systems. Basically, in the US, the horse-trading happens before the election, so you see the coalition that you can vote on at the time of the election. The parties in a parliamentary system with many parties are maybe more analogous to the caucuses. So, we don't have a "party for black people" in the US...but we do have the Congressional Black Caucus, which (mostly) operates inside the big-tent Democratic party.

The fact that parties expect to likely have to throw out some promises in a parliamentary system also comes with some issues. The UK uses FPTP rather than proportional representation, so tends away from having coalitions, but is a parliamentary system, and can do so. It is very likely, from what I've read, that the reason that the UK Brexited was because of some jiggery-pokery associated with this. Basically, the Conservative Party in the UK had promised its voters a referendum on UK membership in the EU. However, at the time this promise was made, the Conservative leadership expected not to be able to achieve a majority, that they would have to form a coalition with the Liberal Democratic party, as they had previously. The Liberal Democratic party was strongly in favor of being in the EU, and probably would have required them to not hold such a referendum as a condition of being in coalition. Holding a referendum is not actually something that the Conservative Party likely wanted to actually do. As a result, the Conservatives could make such a promise and get the electoral support from doing so...with the expectation that they would never have to actually follow through on it, because they'd get the opportunity to throw out some of their electoral promises to voters during the coalition-forming process. However, they did better than expected, and didn't form a coalition, and were stuck holding a Brexit referendum. You won't get that in the US, since the executive makes their promises prior to the election.

NGOs, like the EFF or Greenpeace or the like, also tend to play a larger role in the process in the US, which provides for a lot of options as to involvement in advocacy. In Europe, some countries developed "Pirate Parties", political organizations that work something akin to the EFF here (though there's also the EDRi in Europe, it acts as more of a coordinating institution).

One other issue that parliamentary systems run into is that after the election, they have to decide on a coalition to choose the new executive. This usually doesn't take too long, but sometimes the legislators don't agree in the post-election horse-trading process, and the result is that no executive gets chosen (or, in some cases, as in Italy, a "technocratic" executive gets chosen for the public). Belgium and Northern Ireland have recently had extended periods without an executive (which, in their terminology, is "without a government"), which hampers their ability to do much. In a presidential system, after the election, you know who is going to be running the executive.

In the US, the Big Two parties also hold primary elections, which permits you, as a member of the public (usually registered as a voter of that party, though there are even some exceptions to that), to choose which candidates you want your party to run. That isn't a constitutional requirement, and some parties do not do that. However, it's also input that frequently isn't available to the electorate in Europe, where legislative candidates are selected internally by the party.

This is to elect the President. In a presidential system, as in the US, you choose the leader of the executive portion of government separately from the legislative leader. In a parliamentary system, as many countries in Europe use, the public doesn't choose the leader of the executive portion of government. Instead, they just vote for representatives in the legislative portion, and then those legislators form a coalition (if necessary) and choose a leader of the executive (the prime minister). The closest analog to coalition forming in the presidential election is doing exactly what the Greens are proposing above -- having a candidate drop out and endorse another, with the hopes that they can sway their supporters. It's basically what JFK Jr did, for example, with Trump.

While hypothetically the US could form legislative coalitions, in practice, due to the way the US electoral system works, US parties are essentially equivalent to electoral coalitions in parliamentary systems already -- we already form "big tent" parties necessary to control a house. In the US, the closest analog to this sort of thing actually happening after the elections is when you hear about something like "an independent legislator who caucuses with the Democrats". The US also has weak party discipline compared to many countries in Europe, so legislators are much less constrained to vote along party lines anyway.

Different systems, function kinda differently.

Thanks, though only "Cats" was really just put up for the looks; the other three were to show off some functionality to other folks who might be doing local generation themselves, to let them know about functionality that I'd run into that I wanted and give 'em a heads-up as to it being possible (a way to get decent food photography, a way to emulate artist styles again under Flux, and that the ability to generate progressive transformations in Flux exists).

Or for another example:

UI: ComfyUI

Model: STOIQNewrealityFLUXSD_F1DAlpha

An image divided into four sections showing a progression from left to right.

The image shows a full-body image of a girl.

The image is a drawing.

The first section shows a good, innocent character.

Each section moving rightwards shows an increasingly-evil, increasingly-dangerous and increasingly-corrupted character.

Flux is capable of generating progressions between one image and another. I'm not entirely sure how this is internally implemented, though I'd assume that it does something like have some prompt parsing capable of recognizing syntax sufficiently to detect a progression and then do the equivalent of shifting prompt term weights.

I'd seen plugins that can sort of accomplish similar effects in Stable Diffusion, but the fact that Flux -- and derived models, like this one -- permits for simply taking natural-language directions and doing pretty well on them quite impressed me.

One thing that I like about some Stable Diffusion-derived models is that they have been trained on the works of many artists, can replicate styles.

Flux, while rather more-sophisticated in many ways, including the ability to use natural-language image descriptions, has -- probably intentionally -- not been. And while I really like Flux's functionality, I've been kind of unhappy about losing that.

However, there are models derived from Flux -- as the one used here -- that have had that trained back in.

EDIT: Note that this model still can't do some of the things that I've done in the past with Stable Diffusion-derived models trained on a lot of artworks, like these landscape paintings in the style of Casilear, but the fact that it can do something at all is a considerable improvement over the dev version of vanilla Flux. I did read one comment that the stable version of Flux supports artist names, but then you're stuck using their service to do your generation; the stable model of Flux isn't distributed.