Ceci est une ancienne révision du document !
First we should consider why we might want to load/use Forge. Initially important is that Forge offers a speed increase compared with Automatic 1111. It is rather dramatic for middle of the road GPUs with six or eight gigs of video ram with decreasing speed boast as ram increases but even then some modest amount. Regardless of the speed increase, the inclusion of Advanced features and extensions might make you want to use it anyway. The easiest way to install Forge is via Pinokio (https://pinokio.computer/) as discussed in last month’s issue. Note that it requires an Nvidia graphics card and you need to first load Pinokio. Then within Pinokio, simply start up the Discover option and search for the Forge script via the search bar or scroll down to find the installer. You will need to download the git file as named or rename if needed, then install and run it. Without Pinokio it’s a more typical install requiring a download and usually some terminal work. Rather than repeating what others have already done I will point you to two videos which explain Forge install. You might want to view both before actual installation. Some installations do not require an Nvideo graphics card but will also work with AMD. See: https://www.youtube.com/watch?v=gJotSdaMFqw or https://www.youtube.com/watch?v=TatD9zNvhqY&t=18s and/or other websites or videos as instructions come and go although your best bet would be an installation for your OS.
Also needed regardless of the install may be appropriate models. What is required depends on how you are using Forge given it can essentially act like Automatic 1111 with expanded capabilities. I am running it on two machines, one with Linux Mint using an Nvidia GeForce 4080 graphics card with 16 GB video ram installed via Pinokio and another with Ubuntu Studio using an Nvidia GeForce RTX 3060 with 4 GB video ram using a more standard install. The screenshots of three different model options are shown. They are very similar except the top one is the standard Stable Diffusion interface as being run within the Pinokio framework (4080 card), the second is Forge, without Pinokio, as seen using the 3060 card within Ubuntu Studio and the third as Forge (4080 card) within Pinokio. The later is using the Flux interface. Exactly what is Flux? It’s a relatively new model (https://blackforestlabs.ai/) that has recently gotten a lot of press. Not perfect but very impressive in terms of photographic-like images created, especially when derived from a real photograph. Sometimes it’s a little too perfect or small imperfections make you start to wonder if they are really photographs. Nevertheless, it’s good enough to scare some people into questioning what is needed to control this ever more capable avalanche of image related AI technology. I am not sure who will win between those trying to monetize it on one hand, the pro legislation crowd worrying about nefarious uses on the other with the unappreciated, unaffiliated creative uses in the middle that keep making it better.
Using Forge in Pinokio with a Flux model produces a fairly realistic image of a Bigfoot walking through a library under fluorescent lights with light reflection and a realistic floor. You will notice a variety of options within the screenshot both under the image and on the left are available. Just below the Generate button is an empty style bar which gives you access to what seems to be several hundred styles. You could spend significant time just getting to know the interface. It seems overwhelming and one might have to ask whether you need this kind of creative power. For example if you simply need general images which convey an idea, you might find that Fooocus is more than adequate. Next time I will attempt to address some of the features within Forge using the Flux model. However, for those new to this concept you might want to “try before you buy” meaning get a feel for what is possible before you purchase additional hardware or spend time installing software that you may not need. (It’s already too late for me.) A quick look at https://replicate.com/explore will allow anyone to test the possibilities and still be able to walk away without much cost in time or money and other options do exist. Considering how rapidly this area is advancing and the possible legal issues that may soon start to develop, a prudent approach may be appropriate for those finding the technology overwhelming.