Evoke hosts open source AI models on the cloud for devs and businesses developing AI apps. Run open source models like stable diffusion with our API, and browse new models we frequently add in our mission to make AI more accessible for all.
Hey Jasper! CEO of Evoke here. This sounds like a real problem on our end that I'd like to fix. Can you join our Discord and ping me so we can talk about the error? You can also ping me directly at "Milo G#4162" on Discord, or suggest an alternative contact method.
👋 Hey Product Hunt! I’m Richard, the Co-founder of Evoke.
❓ What’s Evoke?
Evoke is a platform hosting AI models for easy API usage.
We just launched our stable diffusion API, which allows devs making AI apps to focus on building their apps rather than focus on cloud hosting, inference, and MLops.
All you have to do is get your API key and you start running stable diffusion immediately with a few lines of code.
For now, we host stable diffusion v1.5, however, we’re moving quickly to have other models + fine tuning your own models with dreambooth in the near future, so stay tuned for updates.
🚀Here are some benefits we offer over other stable diffusion APIs:
- Faster generations. We’ve optimized it so generations are lightning fast.
- Cheap pricing. We’re geared more towards developers and we understand you don’t want to spend a fortune on testing and making large batches of calls, so we’ve kept prices low.
- Pay as you go. As opposed to a credit system or monthly sub, you’ll pay only what you use.
- Easy to understand documentation. As you probably all know, a poorly documented API is most infuriating, so we’ve made our docs clear and easy to understand.
🤖 How it works
1. Create an account
2. Get your API key
3. Call the API for inference with your preferred settings (prompt, step count, CFG scale, etc)
You can see our docs here
🌎Our mission
When we first started making stable diffusion apps, we hit massive roadblocks with getting it set up on the cloud + knowledge of ML, so this gave us the idea to streamline the process for devs.
Overall, we want to make open source AIs more accessible to all, not just those with high end GPUs or ML and cloud infra knowledge.
Thanks to everyone who supported our journey! 🙏
Happy to answer any questions, and please let us know your feedback :)
👉 Website
👉 API docs https://evoke-app.com/docs
👉 Head over to our Discord for feedback, questions, staying tuned for new updates, or just to hang out
Hey Richard,
I have made an account to test the API. The UX is very intuitive! 😍
Some questions:
1. Is it wise to add Paypal next to Strip since you can pay directly after logging in?
2. Would the user be more convinced to try out a model when you show them the output next to the examples of your REST API?
I have tested your API model in Google Colab and got the same in the Jupiter notebook; when I tested your API with each AUTH token, it gave an error response; I have sent you the screenshots on LinkedIn. 😓
I am comparing implementing your API versus the one from open AI; so far, both seem not Python-friendly and built for Javascript. 🙃
Remember speed to value, I should be able to copy past the script in a cloud console, and it should work with my key.🌱
But congratulations on having position one in the morning; good luck with the launch! 🔥
I can't share this with my network at the moment since it doesn't work; when you have fixed that reach out to me on LinkedIn, and I will share your tool there and on Twitter.
@jas801 1. Thanks for the feedback. We'll be adding Paypal soon
2. Good point. We'll add some pictures in the docs
As for the Google Collab, you're supposed to use one auth token. My bad for not specifying. We'll make sure to tell users about that too.
Thanks for the helpful feedback!
@richard_gao2@jas801 Hey Jasper! CEO of evoke here, I'm working on this issue personally. Can you get in contact with me over Discord at Milo G#4162 ? If not, let me know an alternative contact method I can reach you over.
Congrats on the launch Richard and Evoke team! This seems like something developers looking to experiment with visual AI require these days. Most of us don't want to have to deal with the infra needed to set up AI models.
Are you planning to "host" LLMs for NLP outside OpenAI's too?
Also, what's the latency like? I am assuming you only run/provision these models once you receive a request (a la serverless), so could there be cold start problems that would impact latency and end user experience in general?
@ed_forson 1. Yep, we're planning on hosting LLMs in the near future
2. it's around 3s per image for default settings. There is a cold boot, but the server will stay on for a while after your generation, so if you generate lots of images consecutively, it will only take 3s per image since the server won't shut off
You'll also only be charged for generation time, NOT cold boot time
We are planning on reducing cold boost greatly in the near future. And if enough ppl use it, there will be no cold boot since the servers will be on all the time
Really like that your helping make running/maintaining AI models a more seamless task. There’s a lot of overhead involved especially on the MLOps side of things. Does Evoke help with continuous learning and assessing/serving new models to prod?
@cjries We don't have training yet. But we'll have dreambooth integration soon, which will allow users to train their own AI models via fine tuning. So 100% it will be able to serve new models to prod.
Continuous learning, we haven't looked into yet, but it seems like something that might become big in the future, so we'll check it out.
Report
The technical wins are clear, congrats!
But the marketing could be much better:
> “launch AI apps & SaaS 42% faster…”
> “Save 77% on development time…”
> “Develop 2.44 times faster with..”
> “Never spend your … doing … again.”
> “Create AI tools like a pro, without..”
In your description, get clear and precise.
“Faster generations.” How much faster? 1%? 10%? 26%?
“Cheap pricing” why? Benchmark? Calculate it through on a whole development proces
The innovation is not clear on an emotional level.
Remember that most of your investors have much less to little clue about tech specifications.
Hope it helps, cheers for the awesome innovation.
Evoke