Post Ac4il91AWckdUUSuLA by PeterFalkingham@sauropods.win
 (DIR) More posts by PeterFalkingham@sauropods.win
 (DIR) Post #Ac4bZoww6BzENy8fVw by john@sauropods.win
       2023-11-22T18:42:11Z
       
       0 likes, 0 repeats
       
       I cannot get good results out of #StableDiffusion.“A fox crossing a residential street. The fox has a human face. There are autumn leaves on the ground, terraced houses in the the background, and a slight mist.”It's just ignoring most of my prompt (as well as really struggling with what foxes look like). I've tried many iterations and variations, they're all like this.
       
 (DIR) Post #Ac4bgOIzYmSfmEaWtk by john@sauropods.win
       2023-11-22T18:43:24Z
       
       0 likes, 0 repeats
       
       Aiming for something like this:
       
 (DIR) Post #Ac4bh3LuwoAB5BGweO by miekeroth@socialserver.science
       2023-11-22T18:43:34Z
       
       0 likes, 0 repeats
       
       @john similar experience as I have
       
 (DIR) Post #Ac4bo3MdIqbv94we1I by john@sauropods.win
       2023-11-22T18:44:35Z
       
       0 likes, 0 repeats
       
       @miekeroth The image generators are really uneven. Unfortunately, it seems to the ones that actually work are proprietary.
       
 (DIR) Post #Ac4bvbXocTdJBHmBQO by Zeugs@social.cologne
       2023-11-22T18:46:09Z
       
       0 likes, 0 repeats
       
       @john this is quite normal. What were you expecting? Dall-e 3 is a bit better, but in general that's how they do stuff...
       
 (DIR) Post #Ac4byNxHbIChu50x04 by PeterFalkingham@sauropods.win
       2023-11-22T18:46:50Z
       
       0 likes, 0 repeats
       
       @john @miekeroth Bing (free, using /Dall-E) was a bit better, but didn't catch the human face bit:
       
 (DIR) Post #Ac4cBZDr1JuGc43Dl2 by llewelly@sauropods.win
       2023-11-22T18:49:05Z
       
       0 likes, 0 repeats
       
       @john ... and your residential street is populated with insidious doppelgangers of actual trees.
       
 (DIR) Post #Ac4ckHn1pIiugL3hM8 by john@sauropods.win
       2023-11-22T18:55:11Z
       
       0 likes, 0 repeats
       
       @Zeugs Well, my experience of Dall-e even a while ago was considerably better. Here's what it does with the prompt now. It's ignoring the human face part, but obviously these are light-years better.
       
 (DIR) Post #Ac4dcD9zFuRLxcVjlY by etchlings@wandering.shop
       2023-11-22T19:04:56Z
       
       0 likes, 0 repeats
       
       @john what’s the source on the ideal image here? It’s very unsettling.
       
 (DIR) Post #Ac4dd2y4dRoAdjKxPs by Zeugs@social.cologne
       2023-11-22T19:05:10Z
       
       0 likes, 0 repeats
       
       @john yeah like I said it's a bit better. 🤷
       
 (DIR) Post #Ac4dkhZM4xGwacQQKm by john@sauropods.win
       2023-11-22T19:06:34Z
       
       0 likes, 0 repeats
       
       @etchlings It's my painting. https://johnconway.art/vulpes-persona
       
 (DIR) Post #Ac4e24b1ibXmvgUSxs by etchlings@wandering.shop
       2023-11-22T19:09:54Z
       
       0 likes, 0 repeats
       
       @john love the atmosphere and that fox… being. Are you just seeing if you can make processors give you output similar to what you already create?
       
 (DIR) Post #Ac4e7ELUthGCcBgaBc by john@sauropods.win
       2023-11-22T19:10:48Z
       
       0 likes, 0 repeats
       
       @Zeugs Just a tinsy bit!
       
 (DIR) Post #Ac4eLdilLjTXMwRGvg by miekeroth@socialserver.science
       2023-11-22T19:13:25Z
       
       0 likes, 0 repeats
       
       @john hm.
       
 (DIR) Post #Ac4eNP7gazB4fnwDPU by john@sauropods.win
       2023-11-22T19:13:46Z
       
       0 likes, 0 repeats
       
       @etchlings Yeah, essentially. I don't think artists should be ignoring AI, or pinning their hopes on some sort of copyright salvation, so I'm diving in.
       
 (DIR) Post #Ac4eTDNPBJenyo3qtM by Zeugs@social.cologne
       2023-11-22T19:14:32Z
       
       0 likes, 0 repeats
       
       @john depends on the randomly chosen random seed.🤷 Have you retried the same prompt like ... More often?
       
 (DIR) Post #Ac4eePZ6ClNEy2rzTE by etchlings@wandering.shop
       2023-11-22T19:16:44Z
       
       0 likes, 0 repeats
       
       @john it’s a worthy effort. My vague understanding was that most of the “effective” visual ai was being paywalls, but that’s not something I looked deep at.
       
 (DIR) Post #Ac4eiXgjwdb3JsPhvE by john@sauropods.win
       2023-11-22T19:17:29Z
       
       0 likes, 0 repeats
       
       @Zeugs Yeah, I’ve tried the same prompt six times, with slight variations a few more times. The results are quite consistent, and all terrible.
       
 (DIR) Post #Ac4er8yZC7YVHVqyCu by llewelly@sauropods.win
       2023-11-22T19:19:07Z
       
       0 likes, 0 repeats
       
       @john every aspect of this is so much better than even the DALL-E stuff @PeterFalkingham posted; the fox body is more accurate, the leaves look real, the street looks better, the houses look better, and so on.
       
 (DIR) Post #Ac4exYJtsXVjcdt74q by john@sauropods.win
       2023-11-22T19:20:12Z
       
       0 likes, 0 repeats
       
       @etchlings Sure seems to be the case.
       
 (DIR) Post #Ac4f2RaiK7zuCX5Jqq by john@sauropods.win
       2023-11-22T19:21:07Z
       
       0 likes, 0 repeats
       
       @llewelly @PeterFalkingham Yes, but of course it took me several days to paint rather than 30 seconds to prompt!
       
 (DIR) Post #Ac4fOj269lT6fNWBlY by lauraehall@xoxo.zone
       2023-11-22T19:25:04Z
       
       0 likes, 0 repeats
       
       @etchlings @john this is my view, too. I think it’s not going away so it’s helped a *lot* for me to understand exactly what the various engines are (and are not) capable of
       
 (DIR) Post #Ac4fQJgDOBmucB2btw by john@sauropods.win
       2023-11-22T19:25:33Z
       
       0 likes, 0 repeats
       
       Thought I'd try it with a cat, in case it understands cats better than foxes... holy moly now.I do like it more.
       
 (DIR) Post #Ac4fXH8pE3lYlik2pk by Zeugs@social.cologne
       2023-11-22T19:26:48Z
       
       0 likes, 0 repeats
       
       @john since it can't be the AI's fault it must be you(r prompt) 🙃This fact with the consistency is probably the reason they use it for video now. 🤷
       
 (DIR) Post #Ac4h32FuCqPdp1lHPs by futurebird@sauropods.win
       2023-11-22T19:42:27Z
       
       0 likes, 0 repeats
       
       @john  Ah yes the make you so angry with bad drawings that you are motivated to just draw it yourself machine!
       
 (DIR) Post #Ac4il91AWckdUUSuLA by PeterFalkingham@sauropods.win
       2023-11-22T20:01:45Z
       
       0 likes, 0 repeats
       
       @john Maybe the key part of the prompt that was missing was 'human-like arse' :)
       
 (DIR) Post #Ac4iunZD8NIprEzXQO by john@sauropods.win
       2023-11-22T20:04:33Z
       
       0 likes, 0 repeats
       
       @PeterFalkingham
       
 (DIR) Post #Ac4pR61EfHUy4qsr8i by MesozoicMind@sauropods.win
       2023-11-22T21:17:41Z
       
       0 likes, 0 repeats
       
       @john Of course you can't. AI Art sucks as a concept and in execution.
       
 (DIR) Post #Ac4uohoIp5DxiTuadk by karafuto@mas.to
       2023-11-22T22:17:39Z
       
       0 likes, 0 repeats
       
       @john ok that's actually a fun challenge
       
 (DIR) Post #Ac4wmmnHWUm6VAe9OC by john@sauropods.win
       2023-11-22T22:39:56Z
       
       0 likes, 0 repeats
       
       @karafuto th as t last one! Which AI are you using.
       
 (DIR) Post #Ac4x6BI0fnYtyYGfAG by john@sauropods.win
       2023-11-22T22:43:31Z
       
       0 likes, 0 repeats
       
       @MesozoicMind move forward or die. Don’t worry, I won’t be doing something as boring as being a prompt-jockey.
       
 (DIR) Post #Ac51ZnchczBR52KE2i by Zeugs@social.cologne
       2023-11-22T23:27:49Z
       
       0 likes, 0 repeats
       
       @PeterFalkingham @john @miekeroth It's practically another random seed variation of this:https://sauropods.win/@john/111455719352164697Dall-e seems to have seen more residential areas than SD.Maybe, in the end there is not that much differend stuff in there.
       
 (DIR) Post #Ac51ZoR2bpRvb9sQVM by john@sauropods.win
       2023-11-22T23:33:40Z
       
       0 likes, 0 repeats
       
       @Zeugs @PeterFalkingham @miekeroth It also knows the shape of foxes, which SD clearly does not.
       
 (DIR) Post #Ac52De4QZ2zX2IIFBA by Zeugs@social.cologne
       2023-11-22T23:40:51Z
       
       0 likes, 0 repeats
       
       @john @PeterFalkingham @miekeroth Looks like a fox to me. Source stable diffusion. 🦊🤷
       
 (DIR) Post #Ac52JJkP2O7rk7dDpw by john@sauropods.win
       2023-11-22T23:41:59Z
       
       0 likes, 0 repeats
       
       @Zeugs @PeterFalkingham @miekeroth Try it with legs.
       
 (DIR) Post #Ac52RcqDvlmnHJJduS by karafuto@mas.to
       2023-11-22T23:43:25Z
       
       0 likes, 0 repeats
       
       @john first two are made with https://perchance.org/ai-photo-generator and the last one with craiyon.I did modify the prompt a little, I think the third one had something like "a creature with a fox body and human head with blue eyes". I thought if I describe the face in more details it would actually be added to the image. But oh boy that third one is straight from some horror game
       
 (DIR) Post #Ac53ITYTBhijpDDEhM by Zeugs@social.cologne
       2023-11-22T23:52:51Z
       
       0 likes, 0 repeats
       
       @john @PeterFalkingham @miekeroth This turns into: lets Google that for me with Stable diffusion. It just takes at least 60seconds longer than googling. First shot. Fox running no filters got the legs problem. The shape fits better than on your SD stuff.
       
 (DIR) Post #Ac5NyNBTGNlDJNi7A8 by bhawthorne@infosec.exchange
       2023-11-23T03:44:37Z
       
       0 likes, 0 repeats
       
       @john Can’t you ask an LLM to write the prompt?
       
 (DIR) Post #Ac5waH8rZZHO1BjfVo by mike@sauropods.win
       2023-11-23T10:12:27Z
       
       0 likes, 0 repeats
       
       @john I love the one with no fox!
       
 (DIR) Post #Ac5yWkNJ8vmbQswAU4 by john@sauropods.win
       2023-11-23T10:34:17Z
       
       0 likes, 0 repeats
       
       @bhawthorne Yes, I’ve actually thought about that. The LLM would need to know what it is aiming at, so you’d need to hook in an image description AI to describe a model image, and then tell it to turn that into an effective prompt. I don't know whether they know what makes good prompts though, because they avoid training on AI generated stuff I believe.
       
 (DIR) Post #Ac5zB2QBSyLqzBUjmC by mike@sauropods.win
       2023-11-23T10:40:53Z
       
       0 likes, 0 repeats
       
       @john Did you ask for the cat to be levitating?
       
 (DIR) Post #Ac5zBpVWxsk9DxNCPg by Zeugs@social.cologne
       2023-11-22T23:58:04Z
       
       0 likes, 0 repeats
       
       @john @PeterFalkingham @miekeroth In the end the hard truth about AI generation is that you should kill your darlings early. If the model doesn't hook it's pointless. But look at residential cyberfox he also has back legs but here the model can compensate. It relaxes.
       
 (DIR) Post #Ac5zBqTRN8fKDlP2nI by john@sauropods.win
       2023-11-23T10:40:51Z
       
       0 likes, 0 repeats
       
       @Zeugs @PeterFalkingham @miekeroth I'm not sure exactly what you're arguing, to be honest, but my arguments pretty simple, SD seems a long way behind. I need dozens of attempts to get things like a fox without egregious errors, whereas DALL-E doesn't even seem to make mistakes (although it's interesting it still ignores important components of prompts and absolute will not make a fox with a human face!)Quality is not particularly for what I’m vaguely planning to do, so oh well I guess.
       
 (DIR) Post #Ac5zHKY8yiRwwoDbKS by john@sauropods.win
       2023-11-23T10:42:35Z
       
       0 likes, 0 repeats
       
       @mike I did not, it's the same prompt as the earlier one, just with “cat” substituted for “fox”.
       
 (DIR) Post #Ac5zTROtKiHPfafJfk by mike@sauropods.win
       2023-11-23T10:44:51Z
       
       0 likes, 0 repeats
       
       @john Yikes.
       
 (DIR) Post #Ac60Dx0U9aNb13QY1w by mike@sauropods.win
       2023-11-23T10:45:26Z
       
       0 likes, 0 repeats
       
       @john Now realise that when asked for scientific or technical information, LLMs emit similar errors — they're just harder to spot.
       
 (DIR) Post #Ac60DxldKI5rNHUCWG by john@sauropods.win
       2023-11-23T10:53:12Z
       
       0 likes, 0 repeats
       
       @mike I feel like they're more at the DALL-E level for a lot of stuff. Mostly correct but with small or subtle errors, and gaps in capability.I don't care much about the current error rate for stuff I do. It's adequate. I care more about prompt comprehension, which is also hit and miss (but it's probably as good as people on average).
       
 (DIR) Post #Ac65xAlQLqbiU7lVFg by Gustodon@mas.to
       2023-11-23T11:57:26Z
       
       0 likes, 0 repeats
       
       @john No one gets good results from Stable Diffusion, because all of its results are evil as hell.
       
 (DIR) Post #Ac6HqDcaWmtglo5CJU by bhawthorne@infosec.exchange
       2023-11-23T14:10:38Z
       
       0 likes, 0 repeats
       
       @john Someone had a template for getting ChatGPT to generate stable diffusion prompts. Here it is:https://stable-diffusion-art.com/chatgpt-prompt/
       
 (DIR) Post #Ac6LGFwIpdn3LoPZAG by bhawthorne@infosec.exchange
       2023-11-23T14:49:00Z
       
       0 likes, 0 repeats
       
       @john Okay, here’s the best I was able to do: https://www.bing.com/images/create/an-image-of-a-fox2c-no-fox-face2c-woman27s-face-on-fo/1-655f6604b88a41359d5c0774f3cf4922?id=eDMkDgJr4X1wDGPLRUuu1Q%3d%3d&view=detailv2&idpp=genimg&idpclose=1&FORM=SYDBIC&ssp=1&setlang=en-US&safesearch=moderate
       
 (DIR) Post #Ac6LVKeDbnswIU8pHs by john@sauropods.win
       2023-11-23T14:51:44Z
       
       0 likes, 0 repeats
       
       @bhawthorne That's pretty good, the best one yet.Turns out AIs do write good prompts ChatGPT does it when you ask for an image, and in fact that might be going on invisibly with other interfaces too.
       
 (DIR) Post #Ac6QvOqj6oHEd2F1Xs by bhawthorne@infosec.exchange
       2023-11-23T15:52:27Z
       
       0 likes, 0 repeats
       
       @john I like how despite the woman’s face it still stuck a fox nose on the tip of the human nose.
       
 (DIR) Post #Ac77GnpnbeakH4hCQy by BlackPhi@geekdom.social
       2023-11-23T23:26:26Z
       
       0 likes, 0 repeats
       
       @john The thing about Stable Diffusion is that it is not a tool for following a logical set of instructions. It is about associations and links into features of existing images on the internet. In the case of your prompt, the associations attached to 'autumn' and 'leaves' override the associations of 'terraced houses' and 'residential street'. Also, pictures on the internet of foxes with human faces are, as you say above, not common, so SD doesn't have a lot to work from.
       
 (DIR) Post #Ac77GovrWb2HgAXYWW by BlackPhi@geekdom.social
       2023-11-23T23:37:37Z
       
       0 likes, 0 repeats
       
       @john Another thing about SD is that it can be sensitive to your prompt details. There may be few foxes with human faces on the internet, but there are plenty of foxes so it should be able to do a lot better than your outputs. The challenge of using SD as a tool can often be in getting the feel for how it likes to work and going with it - more like sailing than motoring. Then if you get a picture you like with a fox, SD lets you experiment with infilling the face area with a new prompt.
       
 (DIR) Post #Ac77GpfEntKdwtlnFY by john@sauropods.win
       2023-11-23T23:46:55Z
       
       0 likes, 0 repeats
       
       @BlackPhi I guess I'm a little bit surprised that SD doesn't get which bit is the face and composite, like it can do with subject and background. But the experimenting I’ve seen today leads me to think that although AIs have a strong bias toward the conventional, some can be coaxed to do such things. Dall-E was certainly resistant, but can do it.
       
 (DIR) Post #Ac78cVdA4II28ohOHw by BlackPhi@geekdom.social
       2023-11-24T00:01:59Z
       
       0 likes, 0 repeats
       
       @john Basically SD 'knows' about faces because it has 'seen' millions of images tagged as being faces- almost all of them human. It probably has very little reference to fox faces, other than in children's drawings, so it pretty much ignores that part of your prompt.Dall-E is in many ways a rather different beast to SD: easier to control perhaps, but less inclined to come up with interesting and original 'ideas', I think. It's a bit like watercolour vs oil paint, perhaps.
       
 (DIR) Post #Ac8F0zlkSXgDRmjajI by rakyat@hachyderm.io
       2023-11-24T12:48:22Z
       
       0 likes, 0 repeats
       
       @john For a non-paleontologist like me, dare I say the fox in the third pic look a little bit like a… pterosaur?
       
 (DIR) Post #Ac8FChuifeETXqW0VU by john@sauropods.win
       2023-11-24T12:50:32Z
       
       0 likes, 0 repeats
       
       @rakyat Yes, the very short torso will do that!