For anyone who’s interested I just added a new script to the Logik Portal called Salvador. It will allow you to create images based on prompts with Dall-e 2 from within Flame. The images can then be saved and imported into batch or a library. It can also create variations of images created through Dall-e. A API key from OpenAI is required. You can get when you start an account with them:
OpenAI charges for this, but I believe the highest resolution images it creates(1024x1024) are 2 cents a pop as of right now.
Do you think its possible to intergrate Stable diffusion in a similar manor? Ive seen a video of it in Nuke but that was a while ago now and not sure how stable (pun not instended) it was
Look like it’s a crazy add for Flame, can’t wait to see the evolution of this kind of script.
Thank you for doing this, in a few time it’s maybe going to change our workflow or add more flexibility when a paint or extension look hard to do, good tool in the pocket.
I just uploaded a new updated for this script(v1.1) to the Logik Portal. I’ve added in the ability to select an image in the media panel and send to it Dall-e to create variations of the image. If the image has an alpha channel you can now also do inpainting based on the current prompt. Any area of the matte that is black will be inpainted. The image being selected in the media panel can’t be larger than 1024x1024. That’s the largest image Dall-e will currently accept.
I also realized some of the python packages being installed with the script were for the ARM version of Flame only. So it was only working on M1/M2 Mac’s running Flame 2024. This new version should install the proper packages for x86 versions as well now(linux and 2023.x on Mac). If you installed v1 in a x86 version already, the script will delete those old packages and install the correct ones.
Heeeey! Awesome to see a dall-e implementation inside flame!! Unfortunately it doesn’t seem to be able to install all required packages on my Linux box.
Any hints? This is after i already manually installed some python stuff… i actually managed to install openai manually but it doesn’t seem to be recognised…
I finally got myself a linux box here at home to mess around with. It does seem like there are some issues with the script working in linux. I think I have a pretty go idea of what’s happening. I’ll see if I can fix it over the weekend and post a new version.
I just uploaded a new version of the script to the portal which should hopefully fix the problems getting it to run in linux. If you have a chance to install it let me know how it goes.