LTXV: The first generation of the LTX video model
LTXV introduced core creative control and conditioning capabilities. LTX-2 is the current model, built for higher fidelity and production workflows.
Introducing LTXV
LTXV is the foundational model in the LTX video generation family, combining fast inference, strong temporal consistency, and high visual fidelity for text-to-video and image-to-video workflows.
Real-time performance with cinematic quality
Efficient spatiotemporal modeling enables fast generation without sacrificing motion coherence or detail.
Creative control & customization
Precise control over motion and depth, with support for LoRA and IC-LoRA to apply custom styles, brand identity, and motion behavior.
Open-source and scalable by design
A highly optimized, open architecture built to run efficiently on both high-end and consumer GPUs.
Model Evolution
The LTX video models build on each other, offering increasingly advanced capabilities for creative exploration and production-ready workflows.
LTXV
- First-generation LTX video model built for speed, realism, and creative control
- Creative exploration and flexible workflows
- Video-only generation (no integrated audio)
- Creative control through LoRA and editing tools
- Legacy open-source model, available in LTX Studio
LTX-2
- Advanced LTX model delivering synchronized audio–video generation and high-fidelity output
- Production-ready video generation with extended capabilities
- Native, synchronized audio and video generation
- LoRA support with enhanced style features and camera control
- Available model with open source, API access, Playground, and in LTX Studio
Model Evolution
The LTX video models build on each other, offering increasingly advanced capabilities for creative exploration and production-ready workflows.
01
LTXV
- First-generation LTX video model built for speed, realism, and creative control
- Creative exploration and flexible workflows
- Video-only generation (no integrated audio)
- Creative control through LoRA and editing tools
- Legacy open-source model, available in LTX Studio
02
LTX-2
- Advanced LTX model delivering synchronized audio–video generation and high-fidelity output
- Production-ready video generation with extended capabilities
- Native, synchronized audio and video generation
- LoRA support with enhanced style features and camera control
- Available model with open source, API access, Playground, and in LTX Studio
Model Evolution
The LTX video models build on each other, offering increasingly advanced capabilities for creative exploration and production-ready workflows.
LTXV
- First-generation LTX video model built for speed, realism, and creative control
- Creative exploration and flexible workflows
- Video-only generation (no integrated audio)
- Creative control through LoRA and editing tools
- Legacy open-source model, available in LTX Studio
LTX-2
- Advanced LTX model delivering synchronized audio–video generation and high-fidelity output
- Production-ready video generation with extended capabilities
- Native, synchronized audio and video generation
- LoRA support with enhanced style features and camera control
- Available model with open source, API access, Playground, and in LTX Studio
LTXV vs LTX-2: Model Comparison
Compare the foundational creative capabilities of LTXV with the advanced, production-ready features of LTX-2, the model available and supported today.
Generation Flows in LTX-2
Two flows, optimized for different production needs
Fast
Built for speed and tight feedback loops. Choose Fast Flow when rapid iteration matters more than maximum visual detail.
Technical characteristics:
- Resolutions: 1080p, 1440p, 4K
- Duration: up to 20 seconds
- Lower compute load and faster render times
Pro
High-fidelity generation for stable, detailed results. Choose Pro Flow when visual quality and consistency are more important than render speed.
Technical characteristics:
- Resolutions: 1080p, 1440p, 4K
- FPS: 25 / 50
- Duration: up to 20 seconds
- Enhanced detail and stability across extended sequences
FAQs
What is LTXV?
LTXV is an open source AI video generation model developed by Lightricks as part of the LTX Models family. It supports text-to-video and image-to-video workflows and is commonly used in creator and research environments.
Is LTXV open source and where can I download it?
LTXV is available as an open source AI video generation model on GitHub and Hugging Face:
- GitHub: https://github.com/Lightricks/LTX-Video
- Hugging Face: https://huggingface.co/Lightricks/LTX-Video
These repositories provide access to the model code, files, and documentation for local use and experimentation.
Does LTXV support text-to-video?
Yes. LTXV supports text-to-video generation, allowing users to create videos from written prompts using open source workflows.
How do I use LTXV with ComfyUI?
LTXV is supported in ComfyUI Core as well as by the LTX-Video custom nodes and workflows. They can be installed from our GitHub or through the node-manager in ComfyUI.
Did LTXV get an upgrade or a "LTXV 2.0"?
LTX-2 is the next-generation model following LTXV. LTXV is the first video generation model in the LTX family, while LTX-2 is built on a different architecture and designed for higher-fidelity, production-grade video generation.
What are the different LTXV models, and which one should I use?
LTXV is available in multiple model variants designed to balance quality, speed, and VRAM usage depending on your workflow and hardware.
In general:
- 13B models offer the highest visual quality, but require more VRAM.
- 2B models are smaller and faster, making them suitable for lower-end hardware or rapid iteration.
Key model types include:
- Development (dev) models, which provide full feature support, are trainable, and support the highest quality outputs.
- Distilled models, which are much faster. Geared toward high quality though slightly less prompt adherent.
- Mixed workflows, which combine dev and distilled models for a balance of speed and quality.
- FP8 (quantized) models, which further reduce VRAM usage while maintaining similar behavior.
For most users:
- Choose 13B dev for maximum quality and flexibility.
- Choose 13B distilled for faster iteration with good quality.
- Choose 2B distilled for lightweight, fast generation on limited hardware.
Each model variant is supported by specific inference configurations and recommended ComfyUI workflows, allowing users to select the setup that best fits their needs.
Does LTXV have an API?
No. LTXV does not currently have a public API. The model is available as an open source AI video generation model and is intended to be run locally or integrated through open source tools and workflows, such as ComfyUI.
For API-based video generation, Lightricks provides access through other models, such as LTX-2, which is designed for stable, production-ready API usage.
Does LTXV support LoRA training and LoRA-based customization?
Yes. LTXV supports LoRA-based workflows, allowing users to apply style, behavior, or identity adaptations through LoRA models as part of their video generation pipeline. LoRAs can be used with supported LTXV model variants and integrated through open source tools such as ComfyUI.
LoRA support enables faster experimentation and customization without retraining the full model, making it useful for creative and research-focused workflows.