Hey HN, last week I've been building TattoosAI as a little learning project to get comfortable with Stable Diffusion & DALL-E. I'm absolutely shocked by how powerful SD is... Just like how GPT-3 helped copywriters/marketing be more effective, SD/DALL-E is going to be a game changer for artist! |
I think it would be useful for tattoo artists as well. I'd be totally on-board if my artist whipped something along these lines out on their tablet during a consult, we browsed through some designs based on my inputs, and then have my artist modify/add their personal touch.
Agreed with the artist part. My previous startup (headlime.com, acquired by Jasper.ai), was mostly used by copywriters instead of non savvy folks. I genuinely believe that text2image AI will help artists, not hurt them.
I have an open source example using openai https://observablehq.com/@tomlarkworthy/tarot-backend
You should try to host the model yourself.
Using the SD API is great for a POC, but to scale this and take full advantage of the cost benefits, you need to spin up a cloud GPU.
Heck - you can even build an off the shelf PC and pull workloads from the cloud, and then upload the resulting images. That would work well within your traffic needs. Maybe the upfront cost isn't worth it to you, but if you already have the components ...
(I've done this myself. I run a dual cloud / on-prem 3090 cluster.)
Also, the ideas page has lots of weirdly low quality stuff (eg heath ledger joker does not look like heath ledger, baby yoda does not look like baby yoda, Totoro tattoos are generally bad, etc.). And then there are things like this: https://www.tattoosai.com/tattoos/62ecce4651c8873ccf2bbe67
The pricing model of a one time payment is also quite weird... I get that this is a quick project so it's not polished, but as I would not start charging for it.
Regarding the pricing, it uses Stable Diffusion API which costs $0.02 per image. So I have to charge something to not go bankrupt hehe.
Can you talk more about the stack?
EDIT There ya go: https://www.tattoosai.com/explore
Did you finetune the SD model at all? Or are the tattoo sheets generated just by clever prompting?
But it's great at generating ideas for things that you can then touch up or recreate without the usual AI artifacts and tells (unless that's the aesthetic you're going for).
Such things might go viral in a manner that could hurt (or raise awareness .. it's a high risk gamble).
eg: Snapchat removes Maori tattoo filters after outcry
* Just need to harvest your data first.
But then I'm not a tattoo artist and maybe thats exactly the extra bit creativity an artist sometime wants/needs
"A black and white tattoo design of a bouquet of roses, in the old school style"
Basically structure is "A {black and white/color} tattoo design of {user prompt}, in the {style}"
What does your setup look like? What steps did you take to stand this up as a service?
It looks very well done, kudos! What tech stack are you using for the webapp?
It's just a task library (I used Dramatiq) with a worker that connects to it. Then the worker picks up tasks, processes them and uploads the results to R2.
Stack: NuxtJS/Express(Heroku+Google Cloud Workers)/MongoDB/TailwindCSS
From my research in the last couple days, it only seems that PyTorch will work with AMD cards in combo with RocM, and RocM specifically isn't supporting older AMD gpus that you find on Mac laptops from just 2020.
Can you expand on what MPS is?
https://developer.apple.com/documentation/metalperformancesh...