[11:00 Tue,13.September 2022 by Thomas Richter] |
Recently, the text-to-picture AI Stable Diffusion www.slashcam.de/artikel/News/Die-Stable-Diffusion-Revolution--Kostenlose-Bildgenerierung-per-KI-jetzt-auch-auf-dem-Heim-PC.html was (published), with the help of which one can create fantastic pictures by text input oneself on one&s own computer free of charge (assuming a moderately strong graphics card). However, many users have been deterred by the
;) McDonalds in Edo-period Japan Stable Diffusion Installer for Windows and LinuxBut now someone has - as expected - simplified the whole thing by means of a Windows (and also a Linux) installer, now only one click is enough to install the whole Stable Diffusion package (program plus model with a total of 19 GB) on your own PC. The PC should be equipped with a graphics card with at least 6 GB VRAM, otherwise the computation process will take very long (or the resolution of the standard 512x512 pixel images must be reduced).  subway car toilet bowls and urinals, passengers If necessary, images can be generated by CPU only, but then you need patience: the generation of a 512x512 image can take about 19 minutes on a system with a Ryzen 5 2600 CPU. For comparison: on the same system, an image takes only 40 seconds when using an old Nvidia GTX 1060 GPU with 6GB VRAM. ;) The library of alexandria User interface via browserThe whole thing is operated via browser by means of a simple graphical interface, on which not only the text is entered and the resulting images generated by the AI are displayed, but thanks to a small helper it is also possible to choose from various textual image modifiers at once. These determine the basic style of the generated image and range from a realistic photo to watercolor, ink or comic drawings to imitations of the style of famous painters such as Rembrandt or Picasso. ;) Windows version of Stable Diffusion The result can also be influenced by specifying a certain image mood, such as melancholic, sad or cheerful and colorful. A start image can also be entered, which is then to be varied in a certain way using text prompts. In the menu for advanced settings, among other things, the size of the desired image can be set or whether GPU or CPU is to be used for the calculation and it can be determined how many variations are to be generated. ;) Advanced settings To avoid errors the "stable-diffusion-ui" directory with the installation files should be created directly below the root directory (i.e. C: or D:). After clicking on the install file, the download of the actual data will start - if this aborts prematurely, it can simply be restarted by clicking on the install file again to download the rest.  Steampunk City Terrarium Prompt Design: Good pictures by good prompts.However, some practice is required to communicate with the AI - usually it takes a number of attempts to get a desired or particularly nice result. It helps with so-called prompt design - a developing quasi-science of its own, for which helpful tutorials already exist - to first look at a number of images generated by other users and study their text inputs to get an idea of the range of possible inputs that can influence the appearance of the resulting image, for example  here or here.  Dogs playing basketball in the style of Hieronymus Bosch. Stable Díffusion schon mit GPU mit 2.86 GB VRAMVor wenigen Tagen ist auch ein Patch erschienen, der die VRAM-Anforderungen von Stable Diffusion nochmals optimiert hatte, und es ermöglicht 1280x576 oder 1024x704 große Bilder mit nur 8 GB VRAM zu erzeugen - klassische 512x512 Bilder benötigen mit dem Patch sogar nur noch 2,86 GB VRAM. Wir hoffen, daß auch die Windowsversion von Stable Diffusion bald um diesen Patch erweitert wird und so auf PCs mit noch weniger VRAM lauffähig wird(Danke an Hayos)  more infos at bei github.comdeutsche Version dieser Seite: Kostenlose Bildgenerierung per KI: Stable Diffusion jetzt mit einfachem Installer |
 |
|
|