---
title: "The Best IC-LoRA for Reference Sheet Video Generation — LTX 2.3 in ComfyUI"
source_url: https://www.youtube.com/watch?v=jNWZL2azMjs
captured_at: 2026-08-01
author: "AI Jigyasa"
contributor: "ชิดชนก"
platform: youtube
---
# The Best IC-LoRA for Reference Sheet Video Generation — LTX 2.3 in ComfyUI

youtube video by **AI Jigyasa**, 12m56s, uploaded 2026-07-03. URL: https://www.youtube.com/watch?v=jNWZL2azMjs
Tags: ai, artificial intelligence, machine learning, deep learning, neural networks, chatgpt, ai art, future tech, robotics, computer vision, ai ethics, data science, tech innovation, ai revolution, smart technology, ai assistant, tech trends, ai for business, automation, digital transformation, ai in healthcare, ai education, cognitive computing, ai algorithms, predictive analytics, ai research, ai app

## Shared in LINE Note+
- 2026-08-01 22:41 by ชิดชนก

## Description
Workflow-


Paid Workflow with local Asset (reference sheet Generator)-


In this video, I'll show you how to generate AI videos directly from a reference sheet using LTX 2.3 and IC-LoRA inside ComfyUI — no character setup, no separate image-to-video step, just drop in your reference sheet and let the model do the work. This workflow tests and compares the Ingredients IC-LoRA by Lightricks and the MSR (Multiple Subject Reference) IC-LoRA by LiconStudio side by side, so you can see exactly which one holds character consistency better for your use case.
Featuring IC-LoRA Ingredients: 
Featuring IC-LoRA MSR: 
What you'll learn:

LTX 2.3 Audio-Video Model
IC-LoRA Reference Editing
ComfyUI Workflow & Resource Monitoring
LoRA Comparison (Ingredients vs. Lykon)
Refinement & Enhancer LoRAs
Technical Generation Settings
Character Sheet Prompting
Automated Prompt Generation


Model links
Models (dev)

BF16: 
FP8: 
GGUF: 

Models (distilled)

BF16: 
FP8: 
GGUF: 

text_encoders




vae




loras (distilled)



IC-LoRAs
🔹 Ingredients (by Lightricks) — Reference sheet → video consistency

More info: 


🔹 MSR — Multiple Subject Reference (by LiconStudio) — Multi-character reference consistency

M