t

LTX 2.3 ID-LoRA UGC Video Workflow

A professional ComfyUI workflow for creating realistic UGC-style talking head videos powered by LTX 2.3. This workflow is designed to generate short-form videos from a single reference image while preserving character identity and delivering natural speech synchronization.

The workflow focuses on authentic selfie and smartphone-style content, featuring subtle camera movement, natural facial expressions, realistic eye contact, micro facial motions, natural lighting, and the overall feel of footage captured on a phone rather than in a studio.

Features

• Generates UGC-style talking videos from a single reference image.

• Supports ID-LoRA for improved facial consistency and identity preservation.

• Uses reference audio for better voice, rhythm, and lip-sync control.

• Includes a dedicated audio/video latent pipeline for synchronized video and audio generation.

• Powered by LTX 2.3 22B and the Gemma text encoder.

• Applies spatial latent upscaling for enhanced visual detail.

• Includes RIFE and SeedVR2 post-processing for smoother motion and improved final quality.

• Exports the final video through the built-in SaveVideo node.

Ideal Use Cases

This workflow is perfect for creating AI influencers, UGC ads, promotional videos, product reviews, talking head content, character tests, and short-form videos for TikTok, Instagram Reels, YouTube Shorts, and other marketing campaigns.

It is especially useful when you need more than a simple photo animation. The workflow is designed to produce videos that feel naturally recorded, avoiding excessive cinematic styling, overly perfect stabilization, or polished studio aesthetics. The result is authentic smartphone-style content that feels genuine and relatable.

Core Concept

The goal of this workflow is to combine the controllability of AI video generation with the authenticity of real UGC content. It helps create videos that look and feel like they were recorded by a real person using a smartphone, featuring realistic facial expressions, natural speech, subtle camera movement, and consistent character identity throughout the video.