Training and inference codes here: https://github.com/NVlabs/Sana
Note: I still can't find the model on huggingface yet, according to the code, it should be on hugging face under Efficient-Large-Model/Sana_1600M_1024px
 | I made a stable diffusion based photobooth for my University's halloween day event that uses SD Next's (Vladmandic's fork of A1111) API interface and InstructPix2Pix controlnet. It takes a live view from a webcam, and when the foot switch is pressed, it sends the image to the locally hosted api and displays the output. (In theory I could have the API running on a different server but i don't want to waste my time dealing messing around with firewall settings) Since it took me quite a while to understand the /control API and get it working consistently. I thought that it would be helpful if I shared my code for the photobooth, as well as the "hacky" workaround I used to get it to trigger with a foot switch, if it even helps one person who need help with SD.Next's API. (I used an arduino to trigger the footswitch, that is relatively simple; if you struggle with that, i can provide the pin diagram for the arduino, and the arduino sketch file, it sends CTRL-ALT-P whenever it receive a signal from the arduino) https://github.com/martianunlimited/photobooth Don't mind my codes, it's still a work in progress [link] [comments] |
Conectar