My approach: I generate everything in "Fast" mode first to quickly iterate and find the best results, then I selectively upscale only the videos worth keeping by reusing the same seed.
Generation Modalities
📝➡️🎥 Text-to-Video Create completely new videos from scratch using text prompts.
🖼️➡️🎥 Image-to-Video Animate static reference images using text prompts.
🖼️🖼️➡️🎥 First-Last-Frame-to-Video Generates a coherent video sequence that bridges a defined starting and ending image.
🎥➡️🎥 Video-to-Video Generates synchronized audio and speech driven by video visuals. Offers options to pass through or entirely regenerate the source video.
Audio Input Settings
🔇 No Audio Input No external audio file used. The AI generates new audio based on your text prompt.
🔊 Audio Input Upload an existing voice or music file to drive the animation and lip-sync.
🚀 VRAM Optimization & Long Videos
⚙️ Chunking Settings for Longer Videos
Adjust ffn_chunks based on your video length (at 24fps):
10 seconds (240 frames): ffn_chunks=1-2
15 seconds (360 frames): ffn_chunks=2-4
20 seconds (480 frames): ffn_chunks=4-6
25 seconds (600 frames): ffn_chunks=8-10
33 seconds (800 frames): ffn_chunks=12-16
If you encounter OOM (Out of Memory) errors:
Increase ffn_chunks value
Reduce resolution slightly
Description
LTX2.3
FAQ
Comments (17)
can u please do one without subgraph please ,coz when i unsubgraph it messed everything and i cant use ,my main usecase is custom audio to video and try ur video to video to add audio
Thank you for the LTX 2.3 workflow!! Two things: (1) last frame don't seem to have an impact and (2) where should we connect the Lora loader to add another Lora?
(1) For the last frame not working — this is a known installation problem for many users. You need to manually delete and reinstall the node. Remove the existing folder and reinstall it from here: https://github.com/TTPlanetPig/Comfyui_TTP_Toolset — after a clean install it should work correctly.
(2) For adding a second LoRA — look below the prompt box, there are options with a lora_name field. On the left side you'll also find an "Enable LoRA" toggle to activate it.
@petunia866 Many thanks for taking the time for your guidance! For the Lora, I simply added an "Lora Loader" before the CFG in the subgraph!
hi, how do I add another lore? to use more than one
@dav79mail156 For the Lora, I simply ad a "Lora Loader" before the CFG in the subgraph! Chain is not the best way, but it works depending what you are trying to achieve :)
@obamarx IKR, like how long have powerlora loaders been around?
Latest ComfyUI, brand new conda env, This workflow throws an error:
```
MathExpression|pysssss
module 'ast' has no attribute 'Num'
```
I was able to get the workflow to run by editing ./ComfyUI/custom_nodes/comfyui-custom-scripts/py/math_expression.py
and on line 182-3
change:
```
if isinstance(node, ast.Constant) or isinstance(node, ast.Num):
return node.n
```
to:
```
if isinstance(node, ast.Constant):
return node.value
```
But the output is total garbage static and random noise.
Thanks @q5sys I tried this workaround and I was successfully able to produce a video from an image.
Hi guys im new here. How to download this model Gemma 3 12b abliterated from HF? i there's no 1 .safetensors file but 5 pieces of. Could you help me? I dont know how to download it right way.
For DualCLIPLoader in LTX2.3 you need this:
it is also uncensored
@petunia866 thank you very much!
Awesome workflow. Very convenient to use, huge thanks for putting this together and sharing!
What a neat and well designed workflow. Thank you for sharing it.
Hey question!
What is the chance on the newer version to have a negative prompt box as well?
This workflow uses the distilled model version, which does not support negative prompts.