search

CompVis / stable-diffusion

A latent text-to-image diffusion model
https://ommer-lab.com/research/latent-diffusion-models/
Other
67.91k stars 10.12k forks source link

GUI / notebook for generating images #8

Open cpacker opened 2 years ago

cpacker commented 2 years ago

A graphical interface for generating images (both specifying inputs and visualizing/saving outputs) would make it a lot easier to quickly iterate over prompts and other settings.

For anyone wanting something along these lines right now, I made a basic notebook version of txt2img.py that exposes the args and their defaults (and keeps the model loaded indefinitely): https://colab.research.google.com/github/cpacker/stable-diffusion/blob/interactive-notebook/scripts/stable_diffusion_interactive_colab.ipynb

Basically just change settings on a GUI (e.g., just edit the prompt) and re-run a cell: image

Given the amount of people interested in using the repo, IMO it would be nice to have "official" notebook versions of the main scripts (both img2img.py and txt2img.py) that expose all the options in a similar way. Not sure if the notebook I linked is the best/cleanest way to do it though (e.g. in its current construction, if anyone changes the args in txt2img.py the notebook will go out-of-sync).

mlaprise commented 2 years ago

Using a notebook is ok, but it get out of hand pretty quickly. I don't have access to weights yet, but I'll probably build a small Streamlit app. I'll share it here when it's done.

cpacker commented 2 years ago

@mlaprise another option is to try forking https://github.com/saharmor/dalle-playground for use with Stable Diffusion

0xBLCKLPTN commented 2 years ago

114 I created an Issue where I showed my gui using pyqt5 ;)