Man, you guys are really doing God's work with these! I never thought I'd be checking the AI forums with as much excitement as the main 'IRL' WAM content, but here we are. Strange but exciting times we live in. When we get text to video as realistic as this I don't think I'll ever leave the house again haha
Hi Guys, I thought I would upload a few of my efforts done with Shutterstock powered with Dall-E 3.
I have been following the threads here for getting the prompts right, credit to thereald for suggesting that a gunge tank is just a glass cubicle, and Born Sloppy's Gunge Tank prompt. I have also tried uue404's before and after prompt. I hope it's OK to combine these ideas with some of my own.
mungochutney said: Hi Guys, I thought I would upload a few of my efforts done with Shutterstock powered with Dall-E 3.
I have been following the threads here for getting the prompts right, credit to thereald for suggesting that a gunge tank is just a glass cubicle, and Born Sloppy's Gunge Tank prompt. I have also tried uue404's before and after prompt. I hope it's OK to combine these ideas with some of my own.
I'm curious about your experience with Shutterstock AI. It uses Dall-E 3 just like Bing and Microsoft Designer, so I'm familiar with the quality of it's outputs. But I just read an article that says it has provided tools included to help you creat the image that you want. That intrigues me, as Bing and MD kind of just have your prompt and that's it. The same article also says that they do not have NSFW filtering, which I also find curious. Have you attempted any kind of NSFW prompts to see what happens? I am also unable to find anything about a daily generation limit. Do you know if one exists?
I signed up with Shutterstock, it's 10 pounds or something for a month so I just did that, then there is no daily generation limit.
Haven't tried any NSFW prompts yet but was just edging towards some cake sitting, so we will see, not much if anything gets blocked.
They have a feature that is in beta where you describe what you want and can tell it what is wrong with results, in theory. It doesn't really work atm but is a very good idea.
Like others I can get it to do gunge tanks sometimes but dunk tanks are hard to get right.
Often when you zoom in faces are not quite right, especially if there is a lot of them, which can be so disappointing, but the tech is developing rapidly.
The other thing with Shutterstock is you can download higher resolution results, 5mb or so instead of 250kb.
mungochutney said: Hi Guys, I thought I would upload a few of my efforts done with Shutterstock powered with Dall-E 3.
I have been following the threads here for getting the prompts right, credit to thereald for suggesting that a gunge tank is just a glass cubicle, and Born Sloppy's Gunge Tank prompt. I have also tried uue404's before and after prompt. I hope it's OK to combine these ideas with some of my own.
I'm curious about your experience with Shutterstock AI. It uses Dall-E 3 just like Bing and Microsoft Designer, so I'm familiar with the quality of it's outputs. But I just read an article that says it has provided tools included to help you creat the image that you want. That intrigues me, as Bing and MD kind of just have your prompt and that's it. The same article also says that they do not have NSFW filtering, which I also find curious. Have you attempted any kind of NSFW prompts to see what happens? I am also unable to find anything about a daily generation limit. Do you know if one exists?
I can confirm that they do have a NSFW filter. I had a play around with it and it was very easy to trigger although I've not had a chance to see how easy it is to bypass. Their response to NSFW raises concerns around any image usage though regardless.
"It looks like some of the words in your prompt (the text description you provided to generate the image) may generate images that it would be difficult to safely license for commercial usage. Please modify your prompt and try again."
Yeah... I figured for 10 bucks it wouldn't hurt to try it out. TBH I have no idea how Mungochutney produced the images that they did. I ran into the filter a lot, but I also got a lot of really crap results. Probably one out of every 8-16 images produced was recognizable as being the Dall-E 3 AI. The rest were really weird images that either looked like pasted together pieces of stock photos, or the people looked smeared, and I don't mean with mess. Like if you printed a photo of somebody and then smeared the ink. After playing with it for a day, I also found that I really disliked their styles options. However, I got censored way more often if I didn't used the style options. I don't know... I guess I have a month to fiddle around with it and see if I can figure something out with it.
On that note, I had Microsoft Designer spitting out 2 images at a time for a good part of the day yesterday. I have never seen it to that before. I thought it was 1 photo each generation by default. No matter the prompt, or how tame it is, I've never gotten more than one image before. I thought that was odd. Then suddenly last night, it started blocking everything. I even tried "a photo of a kitten sleeping" and got blocked. Every single day is a new experience lol.
Enjoying these. Looking back on the first images I've posted, I'm pleased with how we've all improved at our prompt craft.
Keep these going - my personal aim is plausible gameshow images that would be the 'mainstream wam' of my dreams, but even if it's more fantastical, if it's a gameshow and it's gunging and it's AI, it fits in this thread.
A few more gameshow pictures. Trying to get away from the "gunge booth" and a bit closer to a dunk tank, the type you could go swimming in. Not 100% there yet, but I still want to have a go in one of these
GungeDrop said: A few more gameshow pictures. Trying to get away from the "gunge booth" and a bit closer to a dunk tank, the type you could go swimming in. Not 100% there yet, but I still want to have a go in one of these
GungeDrop said: A few more gameshow pictures. Trying to get away from the "gunge booth" and a bit closer to a dunk tank, the type you could go swimming in. Not 100% there yet, but I still want to have a go in one of these
I'm amazed by the consistency and texture of your gunge. Any chance of sharing the part of the prompt that describes the gunge?
GungeDrop said: A few more gameshow pictures. Trying to get away from the "gunge booth" and a bit closer to a dunk tank, the type you could go swimming in. Not 100% there yet, but I still want to have a go in one of these
I'm amazed by the consistency and texture of your gunge. Any chance of sharing the part of the prompt that describes the gunge?
There's nothing special about the consistency of the gunge in there, I think. I used "multicoloured gunge", and mentioned "gunge is pouring" (e.g. "gunge is pouring from above [name]"). I do sometimes get images with weird lumpy gunge, but I throw those away.
Loving the stuff you guys are trying out - saw earlier, or perhaps on another thread, the sound-technician getting her just desserts. Great idea
Have to try not getting stuck in a groove of just refreshing a prompt that has turned out a good result, and try original ideas instead to see what's on offer.
mungochutney said: Just a few 'anticipation' pics, ideas for where to go next.
Just getting into the AI. Only used Bing so far. How do you get the girls to turn around? I've literally had 2 or 3 lucky results showing the backsides but all others are front view only.
mungochutney said: Just a few 'anticipation' pics, ideas for where to go next.
Just getting into the AI. Only used Bing so far. How do you get the girls to turn around? I've literally had 2 or 3 lucky results showing the backsides but all others are front view only.
Here's a few of my results so far.
I always just start my prompt with "back view", like "back view photographic image of....". However, Bing doesn't like it. If I want to get a backside shot, I usually use Designer.