For the last couple of days I’ve been testing AI FLUX image generation with exceptional results. It’s hard to believe how good it is, specially compared with StableDiffusionXL, it’s a quantum leap.
So far I start from the torch-vision container:
Step 1:
jetson-containers run $(autotag torchvision)
Step 2:
cd home && git clone https://github.com/black-forest-labs/flux
cd home/flux
pip install -e '.[all]'
I don’t know exactly what you meant. The link you provided is to VoiceCraft, a speech generation synthetizer.
I assumed that you want that I try to start FLUX installation from the stable-diffusion container. I’ve tried that and the response is the same. Around 2 hours until everything is up and ready, because the training data that needs to be downloaded for FLUX to work has nothing to do with StableDiffusion.
If you meant that I should try stable-diffusion. I’ve already tried that, I have been working with StableDiffusion for the last 3 months. StableDiffusion open source delivers extremely poor results. On the other side, FLUX delivers results with a quality comparable to Dall.e 3.
I would rather keep this conversation focused in creating a new docker container. I only need to have a jetson-container for FLUX the same way there is a jetson-container for stable-diffusion-webui.