Show HN: An API for Pix2Pix: Edit Images with Text Prompts (leapml.dev) 26 points by claudfuen 3y ago ↗ HN
[–] thealexshaq 3y ago ↗ Ohhh this is fire. I actually used this api to build https://www.magicstory.ai haha
[–] juliana42 3y ago ↗ Just tried it, love how simple it is, definitely will save a ton of time! Going to use in my next build [–] claudfuen 3y ago ↗ Thanks Juliana!
[–] cloudking 3y ago ↗ Assume this is Stable Diffusion on the backend? Can you tell us about the tech stack and server specs? [–] claudfuen 3y ago ↗ Yes SD on the back-end, currently 1.5 but will also add support for 2.1 soon.We're trying different configurations on the hardware side, currently running on A100s via AWS, but will upgrade this soon to get faster inference/training. [–] cloudking 3y ago ↗ Nice! Are you profitable? Those A100s must not be cheap.I also appreciate you having a free tier, so many other solutions don't. [–] claudfuen 3y ago ↗ Ofc! Surprisingly yes, we do have healthy margins on top of the base cost.
[–] claudfuen 3y ago ↗ Yes SD on the back-end, currently 1.5 but will also add support for 2.1 soon.We're trying different configurations on the hardware side, currently running on A100s via AWS, but will upgrade this soon to get faster inference/training. [–] cloudking 3y ago ↗ Nice! Are you profitable? Those A100s must not be cheap.I also appreciate you having a free tier, so many other solutions don't. [–] claudfuen 3y ago ↗ Ofc! Surprisingly yes, we do have healthy margins on top of the base cost.
[–] cloudking 3y ago ↗ Nice! Are you profitable? Those A100s must not be cheap.I also appreciate you having a free tier, so many other solutions don't. [–] claudfuen 3y ago ↗ Ofc! Surprisingly yes, we do have healthy margins on top of the base cost.
[–] will_wright 3y ago ↗ FYI @claudfuen "Get Started" link in the footer is broken -- redirects to https://www.leapml.dev/singup (typo) [–] claudfuen 3y ago ↗ Ah thanks! Fixing now.
10 comments
[ 3.8 ms ] story [ 26.1 ms ] threadWe're trying different configurations on the hardware side, currently running on A100s via AWS, but will upgrade this soon to get faster inference/training.
I also appreciate you having a free tier, so many other solutions don't.