Home
Creators
Search Recent Random
Posts
Search DMs Popular Hash Lookup Tags Random
Importer
Import FAQ
Register Favorites Keys Review DMs Logout
Community
Telegram ThePornDude
Home Creators Posts Import Register Favorites Logout

1-Click INSTALL Ai-TOOLKIT + COMFYUI Z-Image Turbo LORA TRAINING! (Patreon)

Published:
2025-12-13 17:27:11
Edited:
2025-12-23 00:26:19
Imported:
2026-06
Tags:

Downloads

Content

Hey everyone! I've created a 1-click installer for the “ULTIMATE FREE LORA Training Z-IMAGE-TURBO.” and it’s absolutely insane on low VRAM. This really is the best Ai model currently available and Ai-Toolkit makes training a Lora super ultra easy as well! This lets you train your own LoRAs for the new image king super easily: faces, characters, outfits, full art styles, all with crazy detail on tiny GPUs… it’s just nuts!You can check out the video right here: https://youtu.be/WA5xo7PLDsgI also created a workflow for the XY plot lora comparaison for COMFYUI for my Patreon supporters for Local AND Runpod, that you can find right here: https://www.patreon.com/posts/145769537The installer of course automates the entire install process!

AI-TOOLKIT:

Local install:

Requisites:

– CUDA 12.8 (RTX-50) OR CUDA 12.6 (RTX-40/30/20/GTX)

– Git, Python 3.10 64-bit, Node 18/20 LTS

1. Download the AI-TOOLKIT_AUTO_INSTALL.bat and place it in a folder WITHOUT any spaces in the path!!!

2. Run the bat file and select either option 1 or 2 depending on your GPU

3. Wait for the install to finish

4. ???

5. Profit 😎Use then one of the launcher once you have closed the ui and you want to restart it.

RUNPOD:

1. Create an account if you haven't already: Runpod

2. Click on Pod (on the left side) then click deploy

3. Choose a 4090 or 5090 GPU and chose the pytorch 2.8.0 then edit the template and choose 100GB for both the container and volume disk, then under expose ports, edit:

8888 into 8888, 8675. Then go into environment variables below, click add and then type:

AI_TOOLKIT_AUTH on the left then type a password on the right

then scroll then and deploy.

4. Go to my pods, wait for everything to finish and then click "connect", then "Connect to HTTP SERVICE port 8888"

5. Then drag and drop the AI-TOOLKIT_AUTO_INSTALL-RUNPOD-V2.sh file on the left side of the UI then click on the "Terminal" icon on the right side on the UI

6. Copy and paste these two lines then press enter:

chmod +x AI-TOOLKIT_AUTO_INSTALL-RUNPOD-V2.sh

./AI-TOOLKIT_AUTO_INSTALL-RUNPOD-V2.sh

7. Wait for everything to be installed

8. Go back to the port 8675 and click on it to launch the webui then enter your password you created previously

IF YOU WANT TO RESTART THE UI AFTER STOPPING AND RESTARTING THE POD

Use the same

chmod +x AI-TOOLKIT_AUTO_INSTALL-RUNPOD-V2.sh

./AI-TOOLKIT_AUTO_INSTALL-RUNPOD-V2.sh

command as you did for the installCOMFYUI:LOCAL INSTALL!!! RECOMMENDED TO START FROM SCRATCH

Download the XY-PLOT-LORA-Z-IMAGE-TURBO-ULTRA-COMFYUI-MANAGER_AUTO_INSTALL.bat

Run the bat file

Wait for the install to finish

The webui will launch, load the XY-PLOT-LORA-Z-IMAGE-WORKFLOW.json file and there you go😎

DON'T FORGET TO GO UPDATE COMFYUI BY GOING INSIDE THE UPDATE FOLDER AND RUNNING THE COMFYUI UPDATE BAT FILE!!!!

IF YOU ALREADY HAVE COMFYUI INSTALLED USING A PREVIOUS INSTALLER:

Download the XY-PLOT-LORA-Z-IMAGE-TURBO-MODELS-NODES_INSTALL.bat and place it in the ComfyUI_windows_portable\ComfyUI folder

Run the bat file

Wait for the install to finish

???

Profit 😎

IF YOU GET AN ERROR WHEN LAUNCHING COMFYUI AFTER USING THIS INSTALLER SAYING "AssertionError: Torch not compiled with CUDA enabled" it's ok don't panic.

Just go inside the python embedded folder, click on the folder path, type cmd press enter, this will bring the command prompt window and there type:

python.exe -m pip uninstall torch

python.exe -s -m pip install --upgrade torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu128

--->replace cu128 by your cuda version (if you don't know your cuda version just open a new cmd window and type nvcc --version it will then tell you at the end Cuda compilation tools, release 12.8, V12.8.93 or something different like 12.6, in that case just use cu126 instead of cu128)IF YOU ARE USING RUNPOD:

Create an account if you haven't already: Runpod

Click on Pod (on the left side) then click deploy

Choose a GPU with at least 24gb of VRAM (4090 is best), click here for the comfyui template (aitrepreneur/comfyui), then edit the template and choose 80gb for both the container and volume disk, then deploy on demand

Go to my pods, wait for everything to finish and then click "connect", then "Connect to HTTP SERVICE port 3000" and click on the manager to update comfyui and restart it

Then go to the port 8888, go inside the comfyui folder and then drag and drop the XY-PLOT-LORA-Z-IMAGE-TURBO-AUTO_INSTALL-RUNPOD.sh file on the left side of the UI then click on the "Terminal" icon on the right side on the UI

Copy and paste these two lines then press enter:

chmod +x XY-PLOT-LORA-Z-IMAGE-TURBO-AUTO_INSTALL-RUNPOD.sh

./XY-PLOT-LORA-Z-IMAGE-TURBO-AUTO_INSTALL-RUNPOD.sh

Wait for everything to be installed

Go back to the port 3000 and click on the manager and restart it

Load the XY-PLOT-LORA-Z-IMAGE-WORKFLOW.json file and there you go😎

Also use the command:

tail -f comfyui.log

inside the workspace/logs to get the real time logs so that you always know when the install is done. It will then ask you to restart, so click on that restart button, wait for the webui to reboot, click refresh once it's done and there you go.

As always, supporting me on Patreon allows me to keep creating helpful resources like this for the community. Thank you for your support - now go have some fun😉!

Files

Comments

Nathan0490

Is the XY-PLOT-LORA-Z-IMAGE-TURBO-ULTRA-COMFYUI-MANAGER_AUTO_INSTALL.bat in place of the AI-TOOLKIT_AUTO_INSTALL.bat? Or is in needed in addition to it? It's unclear from the new youtube video. Thanks!

Aitrepreneur

Ai toolkit bat file only installs ai toolkit, the xy plot comfyui installs confyui for those who don't have it for some reason so you can run the xy plot workflow

LW

Nice, K :) Just as a matter of interest... is there any info/rumour/speculation on when the full Z image model will be released? Are we looking at days or weeks or months or do we have no idea at all?

Nathan0490

So I followed all of your instructions to a T in AI Toolkit and when I start the job, it just hangs on " Starting job...". An hour later, still on that with 0 out of 3000 steps. 5070ti 12GB. Any ideas on what could be going wrong?

Nathan0490

thank you for clarifying. im still a noob

Aitrepreneur

No one knows for sure, some say in the next 2 weeks, some say in the next few days... TBH I can't wait for it...

Swatman

trying to install ai toolkit, but when I try to open it up "launcher-toolkit" it tells me VENV not found.

Aitrepreneur

any errors during install? also usually this issue is because you don't have python 3.10.11 added to path.

Swatman

The install quits as soon as it finished making the folder. I have all the folders within the AI toolkit except for VENV. How do I add python to the path?

Nathan0490

No error. Says [WORKER] Starting job [hash] on GPU(s) 0 in the cmd window

Łukasz Pelczar

Do I need to install npm before running installer? I have error: Building and starting the AI-Toolkit web UI... 'npm' is not recognized as an internal or external command, operable program or batch file.

Taiga

is it safe to update pip at this point? I recall, long ago (in comfyUI years) doing so broke all kinds of installs previously and we were told to just ignore the errors.

Vincent Nestler

i have exact same issue on a 5070ti 16gb vram. please share the solution.

Vincent Nestler

Here is a thought. On my 5070 and my laptop 5090... job never starts. On my 4090. starts right up. Exact same install. Maybe a blackwell issue?

George Poulopoulos

Hello , i got that error when i start a new job: Running 1 job WARNING:torchao.kernel.intmm:Warning: Detected no triton, on systems without Triton certain kernels will not work WARNING:torchao.kernel.intmm:Warning: Detected no triton, on systems without Triton certain kernels will not work Error running job: No module named 'torchaudio' Error running on_error: cannot access local variable 'job' where it is not associated with a value

LW

On the one hand, this technology is developing at a dazzling speed, on the other hand it's hard to wait when the next obvious improvement is tantalisingly close. I guess we'll just have to breathe deep and count to ten! Looking forward to your one-click installer when it does drop!

Jason

Installed not found after installing the ai toolkit... Venv not found. Everything was installed as per your videos. No errors.

nellz

Qwen is still better... z-image reminds me of SD1.5...

Greg

Open where you installed AI Toolkit (where the folders assets, config, datasets, docker, etc are located). Type CMD in the explorer line and press Enter, then paste the following in the Command window that appears: venv\Scripts\activate You'll now see (venv) in front of the subject line; that's very important. Then paste the following: pip uninstall torch torchaudio torchvision -y Then when that's done, paste the following: pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121 Once that's done, re-run AI Toolkit, you also may have to kill and delete the job you already had running (or trying to run). Make a new job and run it again and now all should work. The issue has to do with incompatibilities with torch and CUDA versions. Also doing this will only affect your AI Toolkit installation, it won't touch your other CUDA/Torch installations like in ComfyUI as long as your AI Toolkit was installed separately. Also I'm not sure if this fix will work with 5000 series since that one is quirky with CUDA versions

Aitrepreneur

I updated the installer with the V2, download it and try that one

Aitrepreneur

I updated the installer with the V2, download it and try that one

Aitrepreneur

I updated the installer with the V2, download it and try that one

John Cannan

My favorite phrase in the entire world is when you say, "You don't have to do anything!"

Seveon

When I try the comparison workflow in Runpod on a 5090 I get the following error: "CUDA error: no kernel image is available for execution on the device CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect. For debugging consider passing CUDA_LAUNCH_BLOCKING=1 Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions"

Seveon

using a 4090 seems to work. Might be something related to the 5090

Aitrepreneur

Yeah I guess sometimes that might be a bit much😅

Aitrepreneur

Because the comfyui template is not made to work with blackwell gpu architecture without some additional instructions hence why I advise most people to just use 4090s, easier this way

Marcbud

This may just be my experience, but I have noticed that, when experimenting between "De-Turbo(De-Distilled)" and "w/Training Adapter", the De-Turbo(De-Distilled) creates deformed images right in AIToolkit with an NSFW sample prompt : "TRIGGERWORD naked, arms and legs spread". Using the "w/Training Adapter" mode works fine, I have repeated this on three different data sets (6 Loras). The corrupted nsfw generations show up later using the lora in Comfy as well. A am using Runpod with the standard Ostris AIToolkit pod to create the loras.

Aitrepreneur

As I said in the video the de turbo model is a special de distilled zit model so the generation you see during training are not representative of the model since the zit model you have inside comfyui is not the same as the de turbo model. The adapter however does use the zit model hence why the images look better during training. However this doesn't matter once you download the loras and use them inside comfyui

Väinämöinen

Here we go again. Another "1 click installer" and another error: **Exception Message:** cannot access local variable 'model' where it is not associated with a value. I like the worksflows when they are working but I've almost never had one that didn't need fixing :( .. And never got any reply on any issue

Spagett !

Installed with v2 and when I run LAUNCHER-TOOLKIT. It runs but stops at [UI] ✓ Starting... [UI] ✓ Ready in 287ms [UI] (node:5048) Warning: `--localstorage-file` was provided without a valid path [UI] (Use `node --trace-warnings ...` to show where the warning was created) and nothing opens

Jez

So the aitoolkit examples are looking great for me but in CUI i'm just getting 5 identical images despite selecting the loras?! Something obvious i am missing?!

Marcbud

I did a bunch of testing, and, both the DT and Training Adapter LoRa versions seem to provide very similar (good) results in ComfyUI. Thank you! Let's hope the full model drops soon!

Aitrepreneur

I never got any dm from you, I can't see all the message comments I get, Patreon is not made for troubleshooting IT issues. Dm me your issues

Aitrepreneur

have you selected different loras? if you select the same lora each time, it will obviously generate the same images each time

Viktor N

hey Aitrepreneur, would it be possible to make workflow with a Img to Prompt for local? I remember that this was possible in automatic1111 but I don't know anymore what it was called. Would be cool. -> found a way, ollama -> qwen3-vl

Miguel Arce

Hey gm, i'm training locally with 12 gb of gpu. but it's normal that it takes 2 hours only generating the 10 sample images and then starts making the steps and it says it's going to take 3 more hours for that?

Alessa

I used the XY-PLOT-LORA-Z-IMAGE-TURBO-ULTRA-COMFYUI-MANAGER_AUTO_INSTALL

Aitrepreneur

it's a litte slow for the image gen but it depends on the parameters and number of steps

Väinämöinen

I really did DM you (maybe it got lost in all the DM's you receive). I also posted/messaged on Discord. I appreciate you were struggling with your health so I stopped writing after a few DM's. It's just very frustrating having to troubleshoot and sort out the issues. As I said I love your workflows.

Aitrepreneur

Oh then sorry i didn't see it, DM me again then

Michael Ang

it keeps saying winget.exe not found - install app installer in Microsoft store. What app?

vr vr

Get stuck on Creating Buckets when i try to train

Aitrepreneur

you probably have an older windows install, you need to install winget (altough it should install it automatically with the V2 installer) but you can install it through the microsoft store like it tells you to

Pak26

After many attempts, Runpod is not working tonight ! I got this error: Collecting networkx (from torch->lycoris-lora==1.8.3->-r requirements.txt (line 5)) Downloading networkx-3.6.1-py3-none-any.whl.metadata (6.8 kB) Downloading networkx-3.6-py3-none-any.whl.metadata (6.8 kB) Downloading networkx-3.5-py3-none-any.whl.metadata (6.3 kB) error: resolution-too-deep × Dependency resolution exceeded maximum depth ╰─> Pip cannot resolve the current dependencies as the dependency graph is too complex for pip to solve efficiently. hint: Try adding lower bounds to constrain your dependencies, for example: 'package>=2.0.0' instead of just 'package'. Link: https://pip.pypa.io/en/stable/topics/dependency-resolution/#handling-resolution-too-deep-errors root@752ab749127b:/# chmod +x AI-TOOLKIT_AUTO_INSTALL-RUNPOD.sh

Jaina Shop

agree, I have already put in the trash bin Qwen :) as quality of Z-image is excellent.

Alex Roca Sánchez

Same issue here, for the first time I tried runpod to try to install the AI-Toolkit training one, and got the same issue trying both with 4090 and 5090. This is the point where it appears: Downloading networkx-3.5-py3-none-any.whl.metadata (6.3 kB) error: resolution-too-deep × Dependency resolution exceeded maximum depth ╰─> Pip cannot resolve the current dependencies as the dependency graph is too complex for pip to solve efficiently. hint: Try adding lower bounds to constrain your dependencies, for example: 'package>=2.0.0' instead of just 'package'. Link: https://pip.pypa.io/en/stable/topics/dependency-resolution/#handling-resolution-too-deep-errors I tried deleting the AI-Toolking folder of the runpod workspace and install it again, but didn't work. Got the same issue

Aitrepreneur

Github has had some issues, right now it is completely down, so I'll check once it's available again and modify the installer if need be

Aitrepreneur

Github has had some issues, right now it is completely down, so I'll check once it's available again and modify the installer if need be

Aitrepreneur

I have now updated the installer, download the V2 version and don't forget to use the following updated command on runpod: chmod +x AI-TOOLKIT_AUTO_INSTALL-RUNPOD-V2.sh ./AI-TOOLKIT_AUTO_INSTALL-RUNPOD-V2.sh

Aitrepreneur

I have now updated the installer, download the V2 version and don't forget to use the following updated command on runpod: chmod +x AI-TOOLKIT_AUTO_INSTALL-RUNPOD-V2.sh ./AI-TOOLKIT_AUTO_INSTALL-RUNPOD-V2.sh

Alex Roca Sánchez

I tried again with the v2 of the AI-Toolkit runpod installer after it got released and now it works, the v2 fixed the issue I had.

Pak26

It was working great yesterday after you updated to V2 but this morning the HTTP Error 404 on port 8675 is back. Thx

Aitrepreneur

I just tried, it still works, you just need to be patient for everything to be installed and the local url to appear

JSentinel

I ran into the exact same issue, was this resolved?

John Henry

I get the following error. OK 2.7.0+cu128 First run: npm install... 'npm' is not recognized as an internal or external command, operable program or batch file. ERROR: Installation failed during Python/Node dependency setup. Common causes: - wrong GPU choice (cu126 vs cu128) - missing CUDA Toolkit 12.8 - antivirus blocking downloads - unstable internet I have a NVIDIA GEFORCE 5070. I selected option 1 when prompted for GPU type. Any ideas? I also have Cuda Toolkit 12.8 installed.

Lance Crossfire

This is great work! Do you have any guidance around how many images should be used for training to get best results? My understanding is that more is better, but only up to a certain point. I tried 40, which seemed to work well, but 100 images seemed to produce worse results. Is there a sweet spot you would recommend?

JpaLVVmI6c8xM00

Is there a guide on how to use a llm to write the picture descriptions? I only previously used Fluxgym, and Fluxgym doesn't work anymore. I'm trying to reinstall it now.

JpaLVVmI6c8xM00

I found TaggUI and it worked for what I needed.

Lance Crossfire

Ahhh... this seems really nice - thanks for the info!

Rodger Hurlburt

I can't get any of the wan workflows to work anymore CLIPLoaderGGUF only 0-dimensional arrays can be converted to Python scalars

H-BOX

Hi !! I'm using your Wan2.2 long length workflow a lot, since it's the best I could find so far... But have you heard about this ? https://civitai.com/models/2261474/long-videos-with-svi-22-pro-wanvideowrapper-workflow?modelVersionId=2545590 It would allow to keep consistancy EVEN if the character goes out of frame from one part to another...It also supposed to smoothen the transition. But it's pretty incompatible with your workflow, I would really love to see you enhance yours with these new principles !! Thanks and cheers !

Steve

maybe when he puts out the Wan 2.5 content he will also do something for creating WAN 2.5 LoRAs? Hopefully something more in depth than fluffing the example config file in the diffusion-pipe repo. But i'll take what i can get RN.

H-BOX

Did he say anything about wan 2.5 ? I thought it won't be open sourced

Aitrepreneur

No news about wan 2.5 (or even 2.6) for now however if the rumors are true we should have a similar model coming in a week

H-BOX

Ow... I'm looking forward to it !!

Bob

Wan2.6 is out please any updates please make awesome work flow.

Aitrepreneur

Model is not open sourced for now unfortunately

llee8820 (edited)

Comment edits

2026-09-08 20:00 RunPod is only a paid service? If I just want to run things locally on my 4090, should I ignore the RunPod instructions and only do ComfyUI? Thank you
2026-01-07 03:19 FYI for future subscribers... AUTO install does not work. I run into this error... ✓ Starting... [UI] ⚠ `devIndicators.buildActivity` is deprecated and no longer configurable. Please remove it from next.config.ts. [UI] ✓ Ready in 681ms [UI] (node:75168) Warning: `--localstorage-file` was provided without a valid path [UI] (Use `node --trace-warnings ...` to show where the warning was created) DM sent. Thank you.

FYI for future subscribers... AUTO install does not work. I run into this error... ✓ Starting... [UI] ⚠ `devIndicators.buildActivity` is deprecated and no longer configurable. Please remove it from next.config.ts. [UI] ✓ Ready in 681ms [UI] (node:75168) Warning: `--localstorage-file` was provided without a valid path [UI] (Use `node --trace-warnings ...` to show where the warning was created) DM sent. Thank you.

Antoinenorthman

Hello, and thank you for your work. My initial results are quite good when I use only my Lora. However, when I combine it with others, the results become disastrous. Do you have any suggestions for improving compatibility?

Lopez M

hey man I've been following you since this whole thing started, sometime 2022 I think, and you are the best resource for the AI stuff. You don't shy away from NSFW and you keep up to date with everything and come up with these workflow, wow just wow man. Thank you.

llee8820

Was your issue resolved? Runnning into simliar problem

Aitrepreneur

Don't worry, I'm working on it, I had to wait for the quant to be released and functional, still waiting for the repo to get updated. I don't want to release a badly functional half-assed workflow like some other people do

Eyes_Wide_Shut

And that is why we love you! Keep up the good work!

乃ㄩ$卄 乃ㄩ!ㄥᗪ丂

I dont find this "XY-PLOT-LORA-Z-IMAGE-WORKFLOW.json" file or is this the same file Eyes_Wide_Shut is talking about.

Jake

Did he ever respond? This is the error I'm getting too.

llee8820

He did. Sent him my logs and awaiting reply.

Stephan Müller

Ihallo, thanke for your great work! I installed Ostris on runpod but after I stopped and will start again runpods messaging: " Your Pod's GPUs are no longer available. This could be because the GPU hardware has been deployed by another user while this Pod was stopped, or they are undergoing scheduled maintenance". The copy from Runpod (beta version) does not work. Question: Do i have to istall all again on another pod? Every time?? Thanks and greetings.

llee8820

How would I add Z image base to AI toolkit? Rerun the auto install?

BopDobop

The same error... Users are reporting it on his his Discord as well, still unanswered

Complex Avocado

I need my subscription canceled please help. Ive resubscribed on a new account and i’d like to keep subscribing however, i lost my access to my old account 2 months ago, I cant find another way to reach out to you personally. Previous account was Hannaali**995@gmail.com card ending in 3083

Peter J Walker

impossible to find node ImageCropV2! can anyone zip it up for me ?

BadgerWi Johnson

It was working fine until Yesterday and I started getting a kmode_exception_not_handled 0x1e and crashing windows. So I installed the latest NVIDIA driver and then the Tool kit wouldn't load so I Uninstaller and reinstalled. Now it starts and I can create a job but when I click the play button it just spins.

Glen

Can I use this even with the new base model? thanks K