(Workflow Update) LTX-2_I2V (ZIT)(HeartMula)(Reserved VRAM) 20260119 (Patreon)

For Patreon Supporters!

Based on the previous version ( https://www.patreon.com/posts/heartmula-ltx2-148765016 )

Added VRAM management and enhance some setting. As I tested, this workflow can run ZIT , Loras, some smaller tool models, all the way to LTX2 20 seconds length in 1 single RUN without OOM. It allow us to run things in 1 single runtime , efficiency pipeline.

Things you need to know before started it.

Some AI models and tech we have walkthrough previously, if you missed it, here some related posts (include model links in those posts, I won't repeat basic info again) :Z-Image Turbo : https://youtu.be/3mT7KnotPqk

HeartMula Music AI (We use Transcript Audio To Text In This Workflow) https://www.patreon.com/posts/heartmula-music-148593360

The Basic version of this workflow:

https://www.patreon.com/posts/heartmula-ltx2-148765016

Other related examples and video tutorials:

ComfyUI LTX2 + Custom Audio

https://youtu.be/AqyyLY_ajTQ

HeartMula Music AI Model In ComfyUI

https://youtu.be/F9LFAeUbBIs

LTX2 in ComfyUI

https://youtu.be/G2JBqUHUbfQ

Some New Custom Nodes In This Workflow

ComfyUI-ReservedVRAM (This is a new node enable us to Reserve VRAM, so we are able to run ZIT , LTX2 in one runtime)

https://github.com/Windecay/ComfyUI-ReservedVRAM

HeartMuLa_ComfyUI

https://github.com/benjiyaya/HeartMuLa_ComfyUI

ComfyUI-LTXVideo

https://github.com/Lightricks/ComfyUI-LTXVideo

1 - enter your prompt , Bypass last 2 group on the right Group Bypasser, Transcript and LTX2 Sampling wait after ZIT image generated. Or if you have all text prompt in ZIT and LTX2 ready, you do NOT have to use bypass, let the whole workflow run in one runtime.

2- Select Your ZIT Lora, don't touch it if you don't have. Set Reserved VRAM on the right, the Reserved number depends on your GPU, how many GB of VRAM you want to keep without AI consuming. Basically, it keep some memory for yourself, avoid AI consume all and cause OOM.

3 - One you have ZIT image generated, on the Group Bypasser enable audio transcript group , and LTX2 Sampling group.

4 - Config LTX2 Model , High or Low VRAM group depends on your hardware setting

5 - LTX2 Set Your Lora and Reserved VRAM

6 - Set Your Input Audio, keep the ZIT Image input if you were using ZIT group, and your LTX text prompt

7 - Click Run and wait

You will see log , Reserved VRAM before LTX2 stage 1 and stage 1 models active.

8- Done.

Have Fun!