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

NEW ALL-IN-ONE VIDEO TIMELINE EDITOR FOR LTX-2.3! (Patreon)

Published:
2026-05-24 18:56:14
Edited:
2026-06-15 15:55:46
Imported:
2026-08
Tags:

Downloads

Content

EDIT 15 june 2026: right now the ltx node ComfyUI-LTXVideo doesn't work well with the newest comfyui updates. IF YOU HAVE ISSUES, you can use the new FIX-LTXVIDEO-KORNIA.bat file and put it in your ComfyUI_windows_portable folder, next to your run_nvidia_gpu.bat file, run it and it should fix the issues with the node at least until the node receives an update.Hey everyone! I’ve just released a brand new video on the NEW AI VIDEO TIMELINE EDITOR for LTX-2.3, called LTX Director. This is an amazing new addition for AI video generation that basically gives you a full timeline editor directly inside ComfyUI.

You can now control prompts across your video, use multiple frames, create transitions, add custom audio, and direct your generations way more easily than before, all inside one powerful timeline.

So if you wanted more control over your AI videos, this is really one of the best updates we’ve had for LTX-2.3 so far.

You can check out the video right here: https://youtu.be/vfeETjcUpag

The workflow: https://www.patreon.com/posts/159127725

The installer of course automates the entire install process for comfyUI and the models downloads!

LOCAL INSTALL!!! RECOMMENDED TO START FROM SCRATCH

Download the LTX-2-3-ULTRA-COMFYUI-MANAGER_AUTO_INSTALL-V2.bat

Run the bat file

Wait for the install to finish

The webui will launch, load the LTX-2-3_ULTRA_WORKFLOW-V2.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 THEN CHOOSING THE CORRECT LAUNCHER FOR YOUR GPU!!

IF YOU ALREADY HAVE COMFYUI INSTALLED USING A PREVIOUS INSTALLER:

Download the LTX-2-3-MODELS-NODES_INSTALL-V2.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 100gb 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 LTX-2-3-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

Copy and paste these two lines then press enter:

chmod +x LTX-2-3-AUTO_INSTALL-RUNPOD-V2.sh

./LTX-2-3-AUTO_INSTALL-RUNPOD-V2.sh

Wait for everything to be installed

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

Load the LTX-2-3_ULTRA_WORKFLOW-V2.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.IF YOU GET OOM error :Click on the little 3 dots button, arguments variables then enter

EXTRA_ARGS --reserve-vram 10 --cache-nonethen restart the pod

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

TJ

This is great and thank you. Question. Does the timeline mean each clip or prompt are separate generations that are stitched together? Meaning we can have longer videos as long as each segment is under 15 sec? Or is the entire timeline limited to about 15 sec?

Anthony

From my own experiments, 20 seconds is the limit of the model. Still you could use the last frame as a start frame for a new 20 second clip and so on.

Slap Dash Dolt

the lipdub lora is close but I don't think LTX 2.3 is capable of lip syncing a rap verse acapella realistically yet.

rfromp

You said to run LTX-2-3-ULTRA-COMFYUI-MANAGER_AUTO_INSTALL-V2.BAT. But what is LTX-2-3-MODELS-NODES_INSTALL-V2.BAT ? Your instructions just say to run the first one. Do I run the second one after running the first one?

James IV

Hello, I did the 1 click install, activated the relay director, put in a prompt but I have no run button.

BopDobop

How to fix long "Fetch ComfyRegistry Data xx/148" proces whenever ConfyUI starts?

Garion128

Ok, so if i Just use one single prompt everything is allright. But if I try to add an image prompt to a text promp, i get this error : ValueError: There is a segment on the timeline missing a prompt! I checked and there is not a signle part of the timeline missing a prompt... Did i miss something ? ^^" Edit: Ok I was dumb : you need to prompt a text even on your image, that's all =)

V4r0m4st3r

You use auto-install if you dont have comfyui already installed with an older installer. If you have it from an older installer, you move the models-nodes to the ComfyUI folder and run it. This will only install models and nodes, the other one comfy AND models and nodes

V4r0m4st3r

Using the run 16gb file, your workflow only takes up 6gb of vram, but all my 32gb of ram. It is painfully slow. Is there a way to fix this?

DjPan Koumpare

im getting similar error, any ideas? UnetLoaderGGUF Unet Loader (GGUF) ValueError: cannot reshape array of size 24537622 into shape (16384,2304)

DjPan Koumpare

I am getting this Error, maybe someone can help please? UnetLoaderGGUF Unet Loader (GGUF) ValueError: cannot reshape array of size 24537622 into shape (16384,2304)

Paul Mazurkiewicz

Hey if I want to download the Q6 version of the model, where do I get it from? The Q8 is too big for my 24 gb card. I mean, it runs and all but I want faster faster faster!! Great work BTW this model and workflow is INCREDIBLE. Your video is spot on too. Very helpful.

Dirk

Is there a way to automatically save the last frame? In previous versions I used comfyui_last_frame_extractor, but I can't see it anymore in manager. Thanks!

Aitrepreneur

It's still limited to the ltx 2.3 model limits which is max 30s unfortunately but you can use the extend workflow above to increase that

Aitrepreneur

you could use the Get Image or Mask Range From Batch node instead from the kjnodes series

Aitrepreneur

I tried lipdub lora, I took it out of the workflow because I was disappointed by the results, but it depends, I think at higher res lip syncing can actually be pretty good as long as the lyrics are not too fast, I think even seedance have trouble with fast lyrics

Aitrepreneur

it's located usually in the top right corner of the screen, you can move it around wherever you want

Aitrepreneur

This is because of comfyui-manager, not comfyui itself. It is fetching the comfyregistry node database at startup. To avoid that you need to put the manager in offline mode but I don't recommend it obviously

Aitrepreneur

you can choose a higher quant model and not use any vram offload options

Aitrepreneur

This might mean that the model downloaded was corrupted during download, try deleting the model and try the download again

rfromp

Ok, I see, thank you!

Silent Monk

For me its just one vague morphing blobby mess after i rendered a video with the Director using different frames. This is also with the I2V section. Very weird because with your previous workflow V1 it did work. Not sure what I am doing wrong, I selected the same models.

Slava

A runpod question. It takes a lot of time to setup everything, including downloading, etc. (running your script with codex observer). Any best practices?

The Friendly Designer

why do my other comfy ui setup stop working when i try the new setups? how do i make them all stay working?

Aitrepreneur

yeah not normal at all, make sure you got comfyui and all nodes updated if you used the models/nodes only installer. If it's still bad you can dm me screenshots of what you are seeing

Aitrepreneur

Runpod has had some downtime issues lately with slow pods but usuallly setting everything up takes like 15mins on average, update + download, etc

Aitrepreneur

what exactly is the issue? Dm me with more info

The Friendly Designer

i dm you, when i install a new workflow release the other ones give me errors and stop working.

Silent Monk

This is going to be mindblowing, insane, holy shiiit, absolutely amazing, but I had (ofcourse) the wrong lora model selected ;) All good now!

Ameshin

Anyone have any tips for improving generation speed for RTX 3060 (12GB), & 32GB ram for longer videos like 20 seconds using the LTX director? I've been messing around but I'm not sure how to improve the speed.

Nazar Tymtsias

Hi, I tried the install script on a RunPod 4090 instance using your Docker image. Overall it ran end-to-end without crashing, but I ran into two larger pain points that I wanted to flag in case they're useful for hardening the template. 1. Install time is very long (~1 hour cold start) On a fresh pod, the full sequence — venv recreation, Torch cu121 stack, ComfyUI core + manager requirements, custom node clones, custom node requirements, and model downloads — took roughly 60 minutes. Breakdown of what I could observe from the logs: - Torch stack install: a few minutes. - ComfyUI + custom node pip installs (RES4LYF, KJNodes, WanVideoWrapper, Impact-Pack, LTXVideo, Easy-Use, etc.): the longest single block, because each custom node's `requirements.txt` gets resolved and installed individually, with a lot of overlapping deps. - Model downloads from the HF mirror (`Aitrepreneur/FLX`): the GGUF (~22B Q8_0) and Gemma 3 12B text encoder dominate — several GB each, throttled by HF. - Final pin re-application + verification passes. 2. ComfyUI didn't come up cleanly after install A few separate issues stacked on top of each other: - Once ComfyUI was up, the workflow JSON would not queue. Every prompt failed with `missing_node_type` for several node IDs (e.g. `#1648`, `#1806`). The error panel showed the workflow references **core** nodes introduced in ComfyUI 0.22.0, but the install script pins `COMFY_REF=v0.21.1`. Because these are core nodes, ComfyUI Manager can't install them — the only fix was to bump the pinned ComfyUI version. - The error panel also instructed me to run `pip install -U --pre comfyui-manager` and restart with `--enable-manager`, which was misleading: the legacy `custom_nodes/ComfyUI-Manager` was already cloned and loading correctly, and the missing nodes were core, not custom. Environment details, in case helpful: - GPU: RTX 4090 24 GB - Image: based on your `aitrepreneur/comfyui:2.3.5` (or equivalent) - ComfyUI: `0.21.1` (as pinned by the script) - Workflow: `LTX-2-3_ULTRA_WORKFLOW-V2.json` - Model: `ltx-2.3-22b-dev-Q8_0.gguf` Happy to share logs if it helps — let me know which sections would be most useful (install phase, ComfyUI startup, or the workflow load errors). Thanks for the work on the template — once it's running it's great, the friction is mostly in that first hour and the launch handoff.

Morpheus

Is there a possibility to use the "sulphur2Base_dev.safetensors" model? Apparently, it's fully compatible with LTX3.2. It's fully uncensored and trained on human behavior. Could you clarify on how to easily incorporate this into your workflow. This sulphur AI model seem to be incredible. Merci beaucoup ! Just in.. I just discovered they have put out the GGUF model. Thus, it should be easy now.

Bas s

I am getting this error ValueError: There is a segment on the timeline missing a prompt!

Dirk

Is there a detailed guide for lipsync with LTX? In my videos I very often only get a slight camera motion, but no lipsync. Even with combinations of "speaking in perfect lipsync..." and/or "moving his lip in perfect lipsync..." I often get no or wrong lip movement. The highest probability I get with plain frontal faces 2/3 size of image. As soon as the face gets smaller or is half portrait, lipsync will fail. Any hints/guides/loras? Thanks!

Lance Crossfire

@SilentMonk, did you get this sorted out? I seem to be having the exact same problem - "a morphing blobby mess" describes it perfectly, and I also have the previous workflows working fine. All nodes and ComfyUI are fully up to date, and there are no errors when running the workflow.

Silent Monk

@Lance Crossfire yes like I said above, I did not properly select the right LORAs in the LORA loader in each block. They looked selected and there were no errors, but apparently they were not selected. I suggest to double check if the LORAs are actually selected.

uwantmymeat

I have updated everything, fresh install about 5 times and I can not get this to run at all. I get the following: Failed to validate prompt for output 1329:1283: * LTXVLatentUpsampler 1329:1299: - Required input is missing: vae * CLIPTextEncode 1279: - Required input is missing: clip * LTXVSpatioTemporalTiledVAEDecode 1329:1322: - Required input is missing: vae I check that everything is in the correct place and selected. If I click any of the boxes at the top of the workflow and try to run it I get the following: line 847, in throw_exception_if_invalid raise RuntimeError("ERROR: VAE is invalid: None\n\nIf the VAE is from a checkpoint loader node your checkpoint does not contain a valid VAE.") RuntimeError: ERROR: VAE is invalid: None If the VAE is from a checkpoint loader node your checkpoint does not contain a valid VAE. Prompt executed in 0.43 seconds Anyone got any advice? Cheers.

Aitrepreneur

yes ltx 2.3 needs "noise" to generate lipsync, if your audio is too "clean" it will have trouble generating correct lipsync. To solve this, you can take your audio, then record some random background noise, and put them together on top of one another in a software like audacity, then save that audio with audible background noise and use that for the video gen. And if you want to keep your clean audio in the final video, you can use the newest addition in the workflow the audio replacer, where you can easily replace the audio of the video with the clean audio version

Aitrepreneur

this either looks like a corrupted vae model or you are not selecting the correct vae model in the model selection area

uwantmymeat

I figured it out after a few hours lol. For anyone else having the issue you need to lower the version of the sampler node, then it works no issue.

Dirk

So that's the trick!! In fact I worked the opposite way, always tried to use very clean samples so it can "undestand" the speech. No wonder I had problems :-) Makes combined video editing a bit harder since seamless ambient sound is easier to deploy in the video editor as an extra track, but I will get along. Thanks for your support!

Light is Consciousness (edited)

Comment edits

2026-09-05 14:00 I'm getting this erros when I open ComfyUI: 'Missing Node Packs (1) This workflow uses custom nodes you haven't installed yet. To install missing nodes, first run pip install -U --pre comfyui-manager in your Python environment to install Node Manager, then restart ComfyUI with the --enable-manager flag.' Reading the patreon instructions I thought that maybe this could solve it: Type python.exe -m pip uninstall torch in cmd then python.exe -s -m pip install --upgrade torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu128 In the middle of the installation there is a message saying: 'script isympy.exe is installed in '...' which is not on PATH... ' and the installation bars are red and incomplete... I open ComfyUI and still get the same error... So it seems the installation didn't work? I am sorry but I am new to this and this is really confusing... Thanks a lot for any help...
2026-06-05 01:13

Aitrepreneur

dm me but if you got a missing custom node, you can just install it using the manager button in comfyui

WhySoSalty

Getting some errors I think are regarding the Loras or Lora Loader? Power Lora Loader (rgthree) ValueError: buffer length (2001651 bytes) after offset (0 bytes) must be a multiple of element size (2) [ERROR] !!! Exception during processing !!! buffer length (2001651 bytes) after offset (0 bytes) must be a multiple of element size (2) Try switching from the distilled-lora-384-1.1 to just the 384 version and get the following RuntimeError: The size of tensor a (7680) must match the size of tensor b (999296) at non-singleton dimension 2 Tried toggling them both off and got this SamplerCustomAdvanced RuntimeError: The size of tensor a (7680) must match the size of tensor b (999296) at non-singleton dimension 2 Anyone have anything similar happen or possible suggestions to fix/work around?

cleguin

I'm getting the following error trying to run the Prompt Relay Director on the audio VAELoader KJ at the top. I haven't touched anything from default with a clean install and the vae_name is showing the LTX23_audio_vae_bf16.safetensors file. VAELoader KJ RuntimeError: ERROR: VAE is invalid: None If the VAE is from a checkpoint loader node your checkpoint does not contain a valid VAE.

cleguin

After a lot of scouring, I finally figured out it seemed ComfyUI-KJNodes was the culprit. The following at least fixed this for my setup at least. I had to go to the Manager -> find ComfyUI-KJNodes under Installed -> Switch version to 1.4.0.

Luxaeternae

Cant get it to run without generating audio.

Aitrepreneur

in your output folder you get the audio and video only version, so just take the video only version

Mordred Sillence

I followed the Runpod instructions to the T and it didn't work at all! Once I uploaded the install script and ran the commands that you indicated, ComfyUI wouldn't even start afterwards. I am terribly disappointed because I joined the Patreon specifically for this repo.

Aitrepreneur

let me guess...you did everything and after updating comfyui, you couldn't open port 3000? If that's the case, just add a / at the end of the url then press enter, it will open the port 3000 without issues

Rodger Hurlburt

at this point with the video stuff you should just support templates for runpod, like for example you have a whole video on how to tediously train zimage base on loras and yet there is an exisiting ostris toolkit template that I can just load to do the same thing.

Aitrepreneur

I can't know of every template that exists, I just make my own tbh each time and share them here

H-BOX

Hi Aitrepreneur, You made is so clean and easy to use, it's amazing as always. But have you heard about "Licon-MSR" multi reference function ? I think this is really mandatory in this WF so that we can have Maximum consistency on I2V.. Do you think you can update this ???

Aitrepreneur

That is interesting, I haven't tried it yet, for now I can't say since I already have at least 2 videos planned already AND we might have news from the LTX team soon...(maybe a new model?) so we'll see but I'll put it on the list

H-BOX

Thanks for your reply ! I see...Pretty exciting to imagine that LTX might get us an actual UPDATE oO let's see !

RavenWT

Hey there! I don't understand why, but each time I install on Runpod using the .bat, then I restart comfy UI or the pod, it never finishes restarting... and neither comfy on 3000 and jupiter are available. Any ideas?

Aitrepreneur

add a / at the end of the url then press enter

RavenWT

I tried but it doesn't work since it has already a / at the end in my adress bar. I don't figure out why it doesn't work, since the pod says it's ready, but I cannot access it. That's very strange, since I can access Comfy via port 3000 during the install, but as soon the update is done, and I restart it, I cannot access anymore. And I was following the instruction exactly as written in your post. I'm surprised I'm the only one to have the issue with the / Is it a navigator setting?

Aitrepreneur

That is weird indeed and yes try with another browser

Jerald

Followed the RunPod LTX-2.3 install instructions exactly (fresh pod, aitrepreneur/comfyui template, 4090, 100GB container+volume). LTX-2-3-AUTO_INSTALL-RUNPOD-V2.sh ran to completion with no errors — ended with "Install complete / Clean LTX-2.3 RunPod install is ready." After that, port 3000 will not load: Visiting the port 3000 proxy URL gives ERR_INVALID_RESPONSE Adding a trailing / (per your fix in other comments) didn't help Dug into it on the pod terminal and found: nginx is listening on port 3000 (confirmed via ss -tlnp) ComfyUI itself is actually launched on port 3001 by /start_comfyui.sh (--port 3001 is hardcoded in the script) RunPod's exposed/proxied port for this template is only 3000 (and 8888 for Jupyter) — port 3001 isn't reachable externally at all (404 when tried directly) So nginx sits on 3000 with nothing correctly reaching ComfyUI on 3001 behind it I tried manually changing --port 3001 to --port 3000 in /start_comfyui.sh and /app-manager/scripts/start_comfyui.sh, killed the process and relaunched — that just produced Port 3000 is already in use since nginx already owns 3000. So it seems like nginx is supposed to reverse-proxy 3000 → 3001, but something in that proxy config isn't routing correctly. Also noticed a secondary (possibly unrelated) error in the ComfyUI-Manager startup log: alembic.util.exc.CommandError: Can't locate revision identified by '0006_add_loader_path' Failed to initialize database. Can you confirm what the nginx proxy config is supposed to point to, or if there's a step in the docs I'm missing for getting 3000 → ComfyUI working after this specific install script? Happy to share full terminal logs.

Aitrepreneur

It can be a browser issue, try opening the url with another browser in incognito mode for example

Russ Ader

All these bat files are driving me batty. I have an RTX 4090, what bat file should I be using?

Aitrepreneur

I mean read the post. If you don't have a comfyui install already or you want another comfyui install use LTX-2-3-ULTRA-COMFYUI-MANAGER_AUTO_INSTALL-V2 bat file If you have comfyui installed already and just want the models and nodes, use the LTX-2-3-MODELS-NODES_INSTALL-V2 bat file, put that file inside the ComfyUI_windows_portable\ComfyUI folder If you have errors with the ComfyUI-LTXVideo after installing everything, use the FIX-LTXVIDEO-KORNIA.bat file, put it in your ComfyUI_windows_portable folder then run it to fix the issues. Then you can run the normal launcher to launch comfyui

Jose

Hello. Two things before I start: one, thank you for providing all of this, it's great. Two, you helped me previously and I never said thank you, so thank you very much for that. I used the automatic installer and everything downloaded well without a hitch. When it came to launching, I got a couple of error messages. I saw that there's a bat file that could let me update ComfyUI and the Python dependencies, but I'm concerned I will brick my Stable Diffusion WebUI if I do this. It's been bricked twice before, and I really don't want to go through that brand new install process again, if able. I'm sure you'll need to see these error messages to give me a clear, answer, which I can provide, but any help would be appreciated.

Jose

Will do. Let me know if you need anything else.