Stable Diffusion 3 API Now Available
stability.ai
stability.ai
- Free to use for non commercial projects
- 20$/m subscription needed if project has less than $1m annual revenue or <1m users
- Undisclosed "Enterprise" subscription needed for projects larger than above
To me the $240/y sounds pretty awesome for a smaller commercial project, and i like they've set the threshold pretty high up.
But still don't like "undisclosed" black box enterprise pricing though when they're running their own API, because it basically means they can tax you to death if you build a service that competes with their API right - so in effect only viable to create small companies with this tech?
I've seen too many governments and even medium sized companies paying absurd amounts to some shitty locked in cloud platform with lots of better alternatives, because someone got kickbacks, gifts, vacations or a seat on some board, and everyone should be more enraged.
The EU has tried to battle this with various public procurement processes but it's a big clown show of course so i don't know what the solution is.
In many cases I'd even pay a premium to avoid talking to one of those people.
Devs' reaction to all this is honestly the same as when a non-technical exec tells a dev why they "can't just get it done in a day, it shouldn't be that hard right?" People have their own competencies and are usually blind to others'.
I don't talk to sales.
I own the business and I buy stuff all the time, just not from places that say "call for pricing".
I've got too many other things going on to waste time listening to some lame pitch from a guy who was probably selling shoes last week.
If they don't want to sell me stuff because I'm "not their target market", that's fine with me.
I'll buy it elsewhere.
I'm not sure what crazoid B-school theory says that you should make it harder for customers to buy your stuff, but I'm pretty sure that's a really bad theory.
You're continuing to justify the role of salesmen, which is an essentially parasitic role that results in higher prices for everyone.
If my company should get big enough that I have to hire someone to order stuff, I don't want that person wasting time on the phone with salesmen either. Every minute he spends on the phone listening to a sales pitch is a minute he's not actually doing the job I'm paying him for.
"Target market"... pfft.
What was your first clue? The fact that I consider salesmen to be parasites?
>If you don't like the game, don't play it, but don't think you're somehow above the game.
Oh, I am totally above the game, because I will not talk to you or your fellow salesmen. I will not call you. Ever. If you cold-call me, I'll hang up and block your number. If you show up in person, you'll be asked to leave and informed that any future visits will be considered trespass and result in a call to the towing company to have your car removed and a call to the police to have you removed.
I thought that was made clear.
Is this what you call "negotiation"? Because you suck at it, dude. srsly.
You are in the business of conning people into paying a higher price for goods and services for which they would otherwise have paid less. Every penny of your commission is ripped right out of the customer's pocket, for zero value added.
I want products at my doorstep in exchange for $. I don't want a goddamn coffee chat.
How the hell do I know they're not giving me higher prices due to racial profiling or some other unethical reason?
That would allow, at least, some sort of website or chrome plugin to exist that fetches and displays previously-submitted pricing information as an overlay next to any idiotic "call for pricing" statements.
That is the way I would run a country if I was its president.
Maybe there are ways to address the abuse without forcing upfront pricing?
The true costs are not front facing because it may not be clear which part of your product is going to be bearing most of the weight of the enterprise customers use-case until you know exactly what they are buying. Your costs may change a lot based on the nature and scale of what they are buying.
For example, you may be rate limited on a background service they use and the customer usecase will likely pish you over your limit. the next product tier from the background service vendor that's required to fulfill the enterprise company's use-case may cost way more and offer way more capacity than that company is going to use by itself. So you have to make a bet on how much cost to assign to that customer, and how much to assign to other/future customers given you buy that new tier and use it to change what you offer to other customers to try to make use of it. It may lower your cost overall if it supports a feature you can upsell relatively easily. Or it may be a service hardly used at all andyou really can't justify going up to that next tier unless this one enterprise customer pays for nearly all of the added cost
I'm curious why that'd be a concern, since building a service that effectively just resells an API isn't a viable business model.
"undisclosed" doesn't imply changing the price after it was negotiated with you. It just mean they will negotiate the contracts for large projects, without disclosing the terms to the public ... in my view.
or you're in a country with politicians donation/spending limits (which are common outside USA) and they give their software for $1 to "that one annoying politician" and it won't count as a discount towards their donation limit, because the "original price" is unknown.
They could do that anyways. This changes nothing.
Negotiating custom contracts is standard practice at the high end anyways. If your company is spending hundreds of thousands of dollars each month on something at list price, chances are you either got too much money or the company is run by suckers.
I assume stability wants to be in the role of building AI models not applications, so if you provide a clear and nontrivial value add beyond their API you should be fine. If you build a tool that allows architects to make their building renderings more lifelike by adding realistic crowds, vegetation, storefronts etc right from the comfort of their existing tooling it's in Stability's interest that you are successful. If you only resell their API but in green they might not be as invested in your success.
E.g. Spotify: https://www.theverge.com/2023/11/20/23969690/google-spotify-...
So typical "above this negotiating power, let's talk terms", and everybody smaller has to pay list price.
Would save you from the unnecessary calls and coffee chats, you just pay N * subscription fee and get on with your life.
You could probably also use part of the revenue to buy treasury bonds and automagically pay the subscription fees, and it would be a closed intervention-free system.
Maybe the early access beta was limiting the available resources, or they used bad settings, bud juding by that thread it looks like the model got worse during training or the earlier examples were quite cherry-picked.
Agreed that some of the others aren't so bad, however the bar for a convincing swamp creature is a lot lower than for a convincing human being.
If breasts come out looking like that, I'd be taking a hard look at my training data.
If you want the best results there is still skill and work involved. Consequently a showcase by people experienced with the model far surpasses what you get when you shout prompts over the internet for somebody else to try
I am sure someone will tell me there is a reason why I am wrong and these aren't that bad.
Midjourney has never needed an explanation though with words. The proof is in the output. Everything else is nonsense.
Does this mean that SD3 will not be available from Huggingface etc.?
https://www.reddit.com/r/StableDiffusion/comments/1c6awnl/co...
Context: Stability is cutting out thousands of artists for "safety" reason, which means billions of images won't make the cut. I wondered how much the model will be nerfed in terms of nudity, artist names, and the like, and whether these issues can be resolved with some fine-tuning.
Since I still don't have access to SD3, I'm compiling a list of all the Stable Diffusion 3 generations out there. I managed to scrape this info from Twitter using their premium API (which cost me a Benjamin!). I manually curated the images to ensure they are indeed SD3 generations and made it searchable.
In case someone finds it useful, it's available here https://sd3.art/
1. "A stained glass picture of a woman in a library with a raven on her shoulder with a key in its mouth"
2 out of 20 tries
2. "An oil painting of a man in a factory looking at a cat wearing a top hat"
First try
3. "A digital art picture of a child riding a llama with a bell on its tail through a desert"
0 out of 20 tries
4. "A 3D render of an astronaut in space holding a fox wearing lipstick"
2 out of 20 tries
5. "Pixel art of a farmer in a cathedral holding a red basketball"
First try
So about even with these models and much better than previous versions of SD. Better than Midjourney v6.
... Who's supposed to be wearing the lipstick, the fox or the astronaut?
https://www.forbes.com/sites/kenrickcai/2024/03/29/how-stabi...
> By October 2023, Stability would have less than $4 million left in the bank [...] Stability was "underpaying AWS bills for July (by $1M)" and "not planning to pay AWS at the end of October for August usage ($7M)." Then there were the September and October bills, plus $1 million owed to Google Cloud and $600,000 to GPU cloud data center CoreWeave [...] Stability was on track to lose more money per month than it made in an entire year.
This is new, right? All previous releases as far as I can tell were just released on HuggingFace. Does this statement just imply something about licensing for commercial usage, or are they pretending that by gating the download behind a paywall that will somehow require people to buy memberships to get the weights?