I came up with this idea this morning, after 8 hours it's live. It takes any URL/link, downloads the content, finds the inner content, feeds this into OpenAI ChatGPT to create a prompt for Stable Diffusion. Then Stable Diffusion generates an OG/cover image for social media posting.
Hey Everyone - I came up with idea this morning. Wouldn't it be amazing, if you could have OG Image / cover images auto generated for webpages?
Well I built it.
It uses ChatGPT to summarise a webpage, the hard part was finding the right parts of the webpage to include.
Next it feeds this into Stable Diffusion and produces the output.
It's brand new, so will likely have some bugs. I plan to turn this into a paid product once it's working well. Also plan to add a plugin to Wordpress so you can generate a cover image for any blog post.
Lots of ways to improve. But basically if you just drop in the meta tags for every page on your website, it will just work out the box.
Woooo 🥳 Have fun!
p.s. some of the outputs are hilarious. I also include on the website the exact prompt it generated, so you can see how it creating the images.
Cool! I thought how cool it'd be to have such a tool and you made it! (in 8 hours! 👏)
Just a thought: maybe add a default style to the prompt? (Or let the user pick from a list.)
Played with your tool and got "Create an image of a person speaking English with 8 ideas to improve their skills displayed around them.". The result looked like stock photography, sadly a bit cheesy.
Out of curiosity, I asked ChatGPT to generate a description for my article and got:
> A person with a smile on their face, holding a smartphone or laptop, looking excited to talk with an AI language tutor or conversation partner. The background could show a globe or map, symbolizing the idea of exploring new cultures and expanding one's horizons.
Fed it into DALL-E and results were so-so. But if I add a style to try and engineer the prompt then things start to look better. Though it's always tricky. Got decent results with "digital art of portrait, cute pixar chracter, houdini 3d render"
Midjourney supports using images as prompts, so maybe it'd be a better fit?
In any case, it's got great potential, so keep up the good work!
P.S.: by the way, pretty inspiring to see you #BuildInPublic on Twitter!
Convert any URL into a Cover/OG Image
Gliglish
InterSub