Show HN: Stable Diffusion Image-to-Prompt Tool

latentspace.dev

20 points·qasimmunye··7 comments

Given an image, we can search the latent space to produce likely keywords that you can add to your prompt to recreate the target image

Open articleView on HN

Yeah! Img2prompt!

https://replicate.com/methexis-inc/img2prompt

https://colab.research.google.com/github/pharmapsychotic/cli...

And of course, the one built into automatic1111's.

No! Not img2prompt!

These are not at all the same.

Yes, you provide an image, and get prompt(s).

But try out these vs. OP's to see.

You're right. This one is much better and precise with styles and artists and gives you a lot of modifier ideas.

ah, I didn't mean to imply my links were exactly the same technology, merely that the concept of getting the prompt out of an image is a good one.

Naming one thing Image-to-Prompt when there's another thing already called img2promp makes it hard to keep things apart.

How does this compare to "interrogate clip" in automatic1111's webui? More focused on visual styled compared to what the image depicts?

Super interesting project! Is it possible that you can show prompt weights next to the prompt, like how important a prompt is to generate a similar looking image? It would also be nice to see more than 100 prompts.

super interesting. I recently started using stable diffusion images for my blog posts on https://blog.boot.dev . I got really sick of all the same unsplash and pexels images.

Reply on news.ycombinator.com