seems they're overloaded due to popularity
but images are dependent on prompt quality
ChatGPT knows about Xenforo
- Thread starter lucien_lies
- Start date
but images are dependent on prompt quality
"Create a image of a woman and man standing looking at a computer screen during a web design training session. It should be in a business setting. You are viewing them from behind at an angle as they look at the computer screen. The computer screen shows the admin panel for a WordPress web site. The image should be realistic."
What it gave me in return:
The free generators I tried did a much better job.
What it gave me in return:
The free generators I tried did a much better job.
Must of been a bad run. Was that the best variant? Using same prompt with ChatGPT Plus DALL-E 3 I got
Then modified prompt for widescreen image
Yeah not bad. Tried asking it for Centmin Mod logo, but it couldn't find it so gave me this LOL
Prompt engineering will certainly become a valuable skill over time with these tools. There's some great info in the comments section from the ChatGPT and midjourney subreddits to better understand how they achieve certain results. It's hilarious seeing some of the 'jailbreak's' too, that trick the AI.
Yeah, but even using the same prompt, others had better results than I, without the error messages. So clearly it was on their end, not mine.I tried Midjourney free and it did a better job, FWIW.
It appears my initial experience isn't the norm and I'll try again when I have more time (I'm in middle of a couple projects and playing around with the AI's when I wait on batch processes).
Yeah give it another go. This is first time using DALL-E 3 where my prompts produced decent results. Usually I use Playground AI Pro plan with Stable Diffusion XL 2 + DALL-E 2 addon and DALL-E 2 prompts produced poor images compared to SDXL 2.0
Adobe Firefly is now trying the AI thing. It had an issue when I asked it to emulate Picasso, but not Monet.
"Utah Delicate Arch in the style of Claude Monet" (They got the arch somewhat right, but the surroundings are not accurate.)
"Desert road tapering off into the distance"
"Mid century modern room and furniture with wood beams overlooking a cityscape"
It failed completely when I asked it to create a few Nintendo-based characters in various situations. I think they're avoiding that due to copyright panic. It's not that it wouldn't render the characters in a flawed way, but their renderings were not even related to what I asked for.
I also asked it to render a certain hated political person in a rather compromised scenario and it flat out refused!
"Utah Delicate Arch in the style of Claude Monet" (They got the arch somewhat right, but the surroundings are not accurate.)
"Desert road tapering off into the distance"
"Mid century modern room and furniture with wood beams overlooking a cityscape"
It failed completely when I asked it to create a few Nintendo-based characters in various situations. I think they're avoiding that due to copyright panic. It's not that it wouldn't render the characters in a flawed way, but their renderings were not even related to what I asked for.
I also asked it to render a certain hated political person in a rather compromised scenario and it flat out refused!
