2024! Stable Diffusion in Colab Notebook for FREE with no disconnects. FREE Midjourney alternative.

hello everyone today once again we're going to discuss another method of using sa diffusion for free in your lovely Google collab notebook without powerful GPU without any pay subscription and without any disconnects you probably know that I created a lot of different Google collab notebooks for the last 8 months and unfortunately majority of them at this moment is banned so why because they are violating Google color brues and I suppose that only one of them is working is invoke AI but we are aware that it will be eventually be banned so I decided to create a simple notebook for basic uh stable diffusion usage which will work without any bands forever I suppose because its notebook not violating any uh Google color rules so let's get started first of all you need to open the link below under this video and here we should uh choose proper on time press on time here and change on time type here should be Python 3 and T4 GPU then we need to choose the model for our notebook there are two options first one is sdxl and second one ISD XL with refiner important to know that unfortunately is the XL with refiner doesn't work with lower models because of the real restrictions if you don't know what does that mean is the Excel and final here a quick comparation between them as you can see uh thanks to refiner you can get more details on your images and is very important for good generation but at the same time I can say that in majority of cases even uh is the XL without refiner will be quite sufficient for basic usage so let's choose is D Cel and here we should run this cell that's it now we are installing all needed requirements and all needed components and of course sdxl model it takes about 5 minutes here we go we just install all needed requirements and as you can see it has very simple interface but it's a bit cont inuitive I explain why here you should enter your prompt for example a had negative prompt and prompt for example ugly ugly uh image count let it be two you can also enter number here uh style there are a lot of different styles which are available for sdxl anime photography comic books and Etc here also Advanced tab which allows you to enter your seat specify random seat or for constant seat choose your resolution sampler and needed steps for your generation then in order to start image generating you need just press here one button not this one don't run this cell once again you need to run only this cell press generate and that's all it takes about 10 seconds to generate one image and here in a this cell output we will see uh the result here we go we've got our or not ocat important to note that this output is not in full resolution in order to get uh full resolution images you need to open this tab here and here in output images uh all your generate images in Full Resolution just download them by pressing this button that's it there is also a Lowa model is applicable unfortunately I wasn't able to implement this functionality in a bit more better way but at least it works so here you need to enter your link for downloading logo model for example let's take this one from cvti uh press button this button here and copy link address here okay then you need to enter this cell in this uh Link in this cell run this cell in order to load your logo model log loaded very fast and then we need to apply our Lura here you can specify the scale here you should choose your weight name in this case uh the name of this Lura is logo one so let's just run this so so we just applied our logo and we can just generate images with loger here we go we've got our cat with um applied Laura which we just downloaded as I said before ISD XL unfortunately doesn't work with yourin and Laura so probably I will fix that but at this moment it doesn't work if you want to add more models I created another version of that notebook which available for my patreon subscribers that notebook works the same it has the same interface but here you can choose your model delate realistic Vision or custom model which you can specify here I think I also will add as I said before uh image to image Tab and control n but to be honest I just don't know how to implement this in a more understandable way because at this moment it works very very ugly it works stable but looks very ugly so anyway I think this notebook is very good for basic usage but if you want to have access to more sophisticated user interface like band in Google color web UI or probably uh more advanced features in Focus like face WAP or image prompt which is not available in a Google color because of ram restrictions there is only 13 gigs of RAM which is not sufficient for all that features you can get access um to my patreon and also subscribe my channel and me and here I have a lot of different notebooks for different uis which works quite stable in a ugle and Google cop also so that's all for now thanks for watching bye-bye

Channel: marat_ai

Views: 8,421

Keywords: stable diffusion no disconnect, midjourney open source alternative, open-source midjourney, midjourney free unlimited, stable diffusion kaggle notebook, invokeai colab, Automatic1111 kaggle notebook, stable diffusion google colab, stable diffusion free colab, stable diffusion google colab tutorial, stable diffusion, stable diffusion tutorial, stable diffusion lora, stable diffusion free, how to use stable diffusion online, stable diffusion disconnect, Automatic1111 alternatives

Id: 0_9fo7STJBA

Length: 6min 36sec (396 seconds)

Published: Tue Feb 06 2024