Update 4-10-2026: The built-in output from colab no longer seems to work - you get a 403 error when trying to view the ComfyUI page after setting it up, so you'll have to use a 3rd party tunnel. I've provided 2 options in the notebook file (if you select neither, you'll get the built-in output option, which probably won't work).
I'll bet you've been seeing a bunch of hype about this new Z Image Turbo model that just came out, and would love to try it out for yourself? You can run Z image in Google Colab, and it runs VERY well on the instances available on their free tier.
I've included the notebook that will install ComfyUI for you and download all the models you will need to run Z Image Turbo, as well as enabled a ComfyUI extension that will help download LoRAs from civitai.com to further augment Z image's functionality. I've also included a very basic workflow to get you started.
The steps for this will very closely follow my SDXL guide, so you can refer to that one for screenshots if there's something you can't find via just my text descriptions. If you've followed one of my guides in the past, there's really only room in your Google Drive for 1 setup, so it's probably best to either sign up for a second Google account and start fresh, or load your Google Drive and delete your entire ComfyUI folder to start fresh.
Step 1: Get a Google account.
This should be self-explanatory. If you already have a Google account, you can use it, but you are only allowed 15GB total of storage for free, and your Gmail takes up that space, too, so plan accordingly (or sign up for a second Google account).
Step 2: Download workflows and notebook.
Download 'ComfyUI_ZiT.zip' and 'ZiT.json' to your computer (they are both attached to this article). Unzip 'ComfyUI_ZiT.zip' and get the 'ComfyUI_ZiT.ipynb' file out of it. You'll use these to set things up. We'll be saving both of these to your Google Drive, so you'll be able to access them from anywhere in the future.
Step 3: Import notebook and set up Colab.
Go to https://colab.research.google.com. Log in with the Google account you created or chose in Step 1. Choose 'Upload' and upload the ComfyUI_ZiT.ipynb file from Step 2.
There are 3 sections in this notebook. For the first time through, we will run all 3 in order, after that, when you come back to colab later after disconnecting, you'll only need to run the first and third.
'Install Dependencies' section: For the first run through, make sure to check all 4 boxes on the first 'Install Dependencies' section (you can expand a section by clicking the '>' next to it - then you'll be able to see the checkboxes on the right). After your first run-through, you can uncheck the middle 2 checkboxes (uncheck 'UPDATE_COMFY_UI' and 'USE_COMFYUI_MANAGER')
'Download Resources' section: Only run this one time on the first run-through you do. It will save the model, text encoder, and VAE you need to your Google Drive for easy access in the future. I am using GGUF quantizations of these various models here to save on Google Drive space. I have chosen an abliterated version of the text encoder to give you a little more freedom (there's the link to the normal text encoder in that section as well if you want to download it too, you just have to uncomment the line). I'm also using the 'PIG' VAE instead of the image-creator-supplied one, as that is how the default workflow from the gguf creator was set up, and I've found it to work very well.
'Run ComfyUI' section: Once you're all set up (or if this is your second or third or more time through, once you've run the first section), run this to start ComfyUI. It will function exactly as it does in the SDXL article.
Step 4: Load workflow and generate.
Once ComfyUI is up and running, and you've closed out of the workflow templates window and dismissed the banner message about Nodes 2.0 (don't use nodes 2.0 yet), click the 'C' icon in the upper-left, 'File' -> 'Open', and choose the 'ZiT.json' file attached to this article. At this point, you can type a positive prompt into the green prompt box (where it says 'POSITIVE PROMPT') and generate.
There is the same 'Power Lora Loader' node and 'civicomfy' extension installed here as described in the SDXL article, so you can use those to download and load LoRAs. As of this writing, civicomfy doesn't have a good way to search just for Z Image Turbo LoRAs (Z Image Turbo isn't an option for 'base model' in the search screen), so it may be best to search civitai.com directly and then search civicomfy for the exact name of the LoRA.
By default, images will save to the 'output' folder, same as the SDXL setup does.
A note on Google's 'terms of service': Google specifically prohibits using Colab to create 'deepfakes' (although they don't define exactly what they mean by that) in their TOS. Since Z Image is really good at creating realistic images, I thought I would mention that here. (We are already kinda toeing the line of their TOS by using a UI other than the base Colab UI in their free tier, but, yeah).
Have fun with Z Image Turbo!