I've been playing around a lot with Stable Diffusion recently, and I thought it would be fun to showcase this previously impossible use case: business logos that are generated daily.
The implementation itself uses Depth2Img (although ControlNet might be the next step) to generate an image from the logo, as well as some post-processing/masking to clean it up.
I got pretty into Stable Diffusion soon after it came out. Like a lot of users, I tinkered around with different ways to run it, going the usual route of running on my weak local machine, then going on to runpod, then implementing my own custom solution.
What I came up with worked pretty well for me, so I created a site that allows users to upload custom models and run Stable Diffusion “in the cloud”.
I launched in early December and it ended up being more successful than I expected. I just got to $700 MRR, which I’m definitely happy about after years of side projects making exactly $0.
Unless you're wanting people to save the images on the landing page, please optimize the images. WebP and only as big as they need to be rendered.
If I go to a service designed around images and it's taking 5 seconds on a SOLID fiber connection to fully download, it doesn't give me confidence that I'm going to get a fast experience in the rest of your site (even if it's not directly related).
It’s a great point. I had been using BunnyCDN to optimize the images/serve as webp, but there are a few on the model preview page that I definitely need to shrink further.
While we're both here, it's not exactly clear to me what that whole thing means and does. Arguably i'm not too clued up in SD models and what they are and why would I want them. Might be a good idea to explain this or if explanation exists make it more prominent to hook ignorant people like me. :-)
Great point. I might need an entirely separate landing page for the artist/general audience vs the prosumer type landing page that currently exists.
(If you’re interested, the gist is that custom models allow for completely distinct “styles” as well as unique characters. For example, if you wanted to generate art in the style of Monet, you could train a custom model in that style)
The linked post celebrates this saying they can share this data with law enforcement, but is it legal for them to pursue cases based on the evidence found in stolen data?
> But, in some ways, there’s really no such thing as an “illegal” search by a private citizen, at least in the sense that police searches can be illegal: Regardless of issues like lack of probable cause, evidence found by private citizens acting on their own is usually admissible in court. That’s true even if the private citizen committed a crime like trespass or theft to accomplish the search.
It does, but in a murder trial, the weapon was generally out of police custody for at least some time, right?
There's an opportunity to introduce reasonable doubt when a third-party is in possession of the data in-between, but it's likely this sort of data isn't going to be the only evidence in any resulting prosecutions. It's far more likely to be probable cause for warrants.
It does, but it’s actually the fourth amendment that would be in place here, and numerous child pornography cases as well as the Panama papers have shown that US courts will allow illegally obtained data, as long as it was obtained by private citizens who were not working with law enforcement and that the data can be established as reasonably untampered.
She, as most politicians, compete in a primary - that is where the real election is. Her district was going to vote for the Democrat candidate no matter who was selected in the primary. Lastly 57% may seem like a lot or whatever, but it's ~16,000 votes out of a potential ~700,000 for the district. Hell, 16k out of ~100,000 that voted last time isn't a huge amount.
Great work! I just wanted to point out that I'm seeing "Plans start at $xx.95/month" on the homepage. Wasn't sure if that's a typo/something you meant to replace.
Thanks for checking it out :) And I agree with you - it's definitely a small use case. I think it would be most helpful for companies with a small dev team who want to implement a favorites list, but may have other, more pressing things on their backlog.
It's an idea that has been in my head for a while. I like the idea of small feature-based APIs that can be used without requiring much dev time.
If they like it enough to sign up/give you their email, what’s preventing you from sending them an update? Or sending them a simple questionnaire?
I would be careful when considering an upvote a positive signal. At the best, it’s someone showing they are just on the line of being interested enough to sign up. At worst, it’s just empty validation for an idea that won’t work.
