Show HN: TattoosAI – AI-powered tattoo artist, using Stable Diffusion
tattoosai.comHey 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! Giving up my email for a try is a no-go from me, but it's definitely a super interesting concept and, at least from the quick video, I could see this becoming fairly popular if the output is reasonable (and probably an adjustment on pricing -- I would not pay for it as is). 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. I REALLY wanted to allow testing without email. The issue is that every trial costs me money. Not "paywalling" it with a login would mean people can just generate unlimited times.Bad
payoff but had to :/ 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 use an invisible firebase anonymous login for public rate limiting. I have an open source example using openai https://observablehq.com/@tomlarkworthy/tarot-backend A demo sells it. 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.) Planning on learning how to do this next week. If you have any articles/video that could point me in the right direction, that would be highly appreciated. I made this for a generator I made (example, https://phantasmagoria.stavros.io/images/6cJAbYyeDXfA6oHZ/). 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. You’re losing a huge market share by not solving this. I would have generated a few, but I’m not giving you or anyone else my email unless I really need something. Cookies/ip rate limiting. I'd rather sign up with a fake email than potentially be rate limited by someone in my same IP range. Why would they limit an IP range rather than a single IP? Do you live in a country that is an internet pest for these sort of things? Yes. Unfortunately there is not a good solution here. FYI, you still have "Lorem ipsum dolor sit amet, consec tetur adipis elit " in the pricing page. 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. Cheers for the feedback, deploying fixes now! 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. Ah so you're not hosting SD yourself in the cloud, you're just using Dreamstudio's API? Excellent idea, much simpler setup, and $0.02 actually seems pretty cheap. Can you talk more about the stack? Lol I wish I knew how to self host, would probably save me some bucks. Stack: NuxtJS/Express(Heroku+Google Cloud Workers)/MongoDB/TailwindCSS There are many tutorials on how to set this up on AWS but I'm not sure that would be cheaper than what you have now. So if I understand correctly, the web site seems to take input from the users, build a phrase from the selections(like maybe,"A black and white tattoo of a bouquet of roses, in the old school style") and then pass that phrase to stable diffusion and generate the images. If you have access to DreamStudio, I think we can try out phrases like this for free and see the output. It would be nice to explore the catalog of created art. I don't want to register nor generate myself, but I'd like to see the creativity in people's prompt leading to the creativity of the AI. Love this idea. Building that right now. EDIT
There ya go: https://www.tattoosai.com/explore You need a (probably paid) option to keep designs private if you want professionals to use this. MidJourney does this for example as an add-on price. Yeah, have to protect those professional private designs! Otherwise someone might train their AI on them and take your job. Incredible! This is going to make tattoos of Chinese characters even worse than they already are :-D Hey Danny, super impressed with the results. Don't really get all the negativity here. Did you finetune the SD model at all? Or are the tattoo sheets generated just by clever prompting? Thank you! Just clever prompt building. Took many, many hours, but finally find the right fit. Definitely a cool use for this sort of thing. I've been playing with SD for the past couple of weeks on my home desktop. It's not at a stage where you'd necessarily use its output directly as a print or certainly a tattoo on your skin. 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). You might want to avoid any "tribal" knock off tattoos from AI generation .. 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 read that one, I'll start working on a content filter :) Try it out for free*!! * Just need to harvest your data first. Kudos for the idea and putting the service together.
In practice I wouldn't let my tattoo artist ink myself if they couldn't come up with sketches like the ones on the explore designs page. But then I'm not a tattoo artist and maybe thats exactly the extra bit creativity an artist sometime wants/needs The following phrase seems to work in case someone wants to try out directly on SD DreamStudio: "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}" Hey Danny. I just got SD running on my Windows desktop last night and spent a few hours today finding out why it won't run on my 2020 MacBook Pro. What does your setup look like? What steps did you take to stand this up as a service? I rewrote this SDK (https://github.com/Stability-AI/stability-sdk) to Javascript and turned it into a Cloud Worker. Quite the challenge, but I got it to work eventually. The lstein fork [1] of the CompVis main repo is working on "Apple Silicon" based machines (and may work on Intel based too). It's not very fast though, ~3.5 minutes for 50 steps on my 16GB M1 Mini, whereas I understand that a 3080 can spit them out in the 30 second range. M<x> machines with higher GPU core count I would suppose are faster. Intel Macs with AMD GPUs are supported by the MPS backend. Hey you may not get this reply till much later but I'd love more info. 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? MPS is Metal Performance Shaders which is Apple's MacOS libraries for ML workloads. The MPS libraries are only on MacOS, but support both Apple Silicon and AMD GPUs. This means that on MacOS, you specify the 'mps' backend to pytorch as the device instead of 'cuda' or 'cpu', and MacOS runs operations on whatever GPU is available, be it an M1 or an M2 or an AMD GPU. https://developer.apple.com/documentation/metalperformancesh... Presumably they need > 10GB of VRAM though? I'm guessing my 2019 MBP with only 4GB is going to say no. I don't have an AMD Mac to test on but on the Nvidia side of things there's support now for 4GiB cards with the right configuration, so it might be possible. Yeah those tattoos will have a very stable ink diffusion. Why the obsession with the joker in every example page? Some old SEO testing with the previous DALL-E AI. Removing those now. A lot of people on the discord channel of mid journey have been using it to generate artwork that they intend to have tattooed on them. I'd like to try it but when I put in a (valid) email address at my personal domain it says "Please fill in a valid email" Sadly unable to try without an email address, and it won't take my Firefox Relay mail proxy. Pass. Oh no, I got ninja'd, I was just thinking of making something like this myself! It looks very well done, kudos! What tech stack are you using for the webapp? Nuxt, NodeJS (partly Heroku, some Google Cloud workers), MongoDB, TailwindCSS This is awesome nice work Thank you!