# LTX-2 | Lightricks | Knowledge Base | Instagit

Official Python inference and LoRA trainer package for the LTX-2 audio–video generative model.

GitHub Stars: 7.7k

Repository: https://github.com/Lightricks/LTX-2

---

## Articles

### [Multi-GPU Pipeline Runners in LTX-2: Available Runners and Work Distribution Strategies](/Lightricks/LTX-2/what-multi-gpu-pipeline-runners-available-ltx-2)

Discover LTX-2 multi-GPU runners like TI2VidTwoStagesRunner and DistilledRunner. Learn how they use sequence parallelism, tiled data parallelism, and distributed VAE decoding to speed up video generation.

- Tags: deep-dive
- Published: 2026-08-20

### [How Gradient Estimation Reduces Inference Steps in LTX-2 Video Generation](/Lightricks/LTX-2/how-does-gradient-estimation-reduce-inference-steps-ltx-2)

Discover how LTX-2 uses gradient estimation to cut video generation inference steps from 100 to 2. Learn how velocity-corrected sampling maintains quality with fewer steps.

- Tags: performance
- Published: 2026-08-20

### [Convolutional vs Diffusion Video VAE in LTX‑2: Architecture, Speed, and Use‑Case Comparison](/Lightricks/LTX-2/difference-between-convolutional-and-diffusion-video-vaes-ltx-2)

Compare LTX-2 video VAEs: explore ConvVAE vs DiffVAE architectures, speed, and use cases. Understand their encoder and decoder differences for optimal performance.

- Tags: comparison
- Published: 2026-08-20

### [How to Configure Model Offloading (CPU vs Disk) for Memory-Constrained Systems in LTX-2](/Lightricks/LTX-2/how-to-configure-model-offloading-ltx-2)

Configure LTX-2 model offloading to CPU or disk for memory-constrained systems. Run massive diffusion models on limited VRAM by streaming weights from system RAM or NVMe storage.

- Tags: how-to-guide
- Published: 2026-08-20

### [LTX-2 Prompt Writing and Enhancement: Best Practices for High-Fidelity Video Generation](/Lightricks/LTX-2/best-practices-prompt-writing-ltx-2)

Master LTX-2 prompt writing with best practices for clear, concise prompts and automatic enhancement using --enhance-prompt to achieve high-fidelity video generation.

- Tags: best-practices
- Published: 2026-08-20

### [How Duration Prediction Works to Auto-Determine Frame Count in LTX-2](/Lightricks/LTX-2/how-does-duration-prediction-work-ltx-2)

Discover how LTX-2's DurationHead predicts video duration in seconds, automatically setting frame count while respecting VAE temporal constraints for efficient video generation.

- Tags: deep-dive
- Published: 2026-08-20

### [DiffVAE Backend Options in LTX-2: NATTEN, Triton, Eager SDPA & Blackwell DSL Explained](/Lightricks/LTX-2/what-are-diffvae-backend-options-ltx-2)

Explore DiffVAE backend options in LTX-2: NATTEN, Triton, Eager SDPA, and Blackwell DSL. Understand which to use for your hardware and configuration.

- Tags: deep-dive
- Published: 2026-08-20

### [How Generated Keyframes Work in LTX-2 and Their Token Cost Explained](/Lightricks/LTX-2/how-generated-keyframes-work-ltx-2)

Learn how LTX-2 generated keyframes work by adding latent tokens for conditioning frames. Understand their token cost calculated as N / num_latent_frames.

- Tags: deep-dive
- Published: 2026-08-20

### [How the Dub-It Pipeline Works for Audio Rephrasing with Lip Sync in LTX-2](/Lightricks/LTX-2/how-dub-it-pipeline-works-ltx-2)

Discover the Dub-It pipeline in LTX-2 a two stage diffusion system for audio rephrasing with lip sync. Learn how it preserves original audio and maintains perfect sync.

- Tags: deep-dive
- Published: 2026-08-20

### [How to Use IC-LoRA for Video-to-Video Transformations in LTX-2: A Complete Guide](/Lightricks/LTX-2/how-to-use-ic-lora-for-video-to-video-ltx-2)

Master IC-LoRA for video-to-video transformations in LTX-2. This guide details using ICLoraPipeline and VideoToVideoStrategy for seamless style and structure transfer.

- Tags: how-to-guide
- Published: 2026-08-20

### [LTX-2 Guidance Parameters: CFG, STG, and Modality Guidance Explained](/Lightricks/LTX-2/what-guidance-parameters-does-ltx-2-support)

Master LTX-2's guidance parameters: CFG, STG, and modality guidance. Tune these signals independently for advanced video and audio generation. Learn more now.

- Tags: deep-dive
- Published: 2026-08-20

### [How to Apply LoRA Adapters (Distilled, IC-LoRA, Detailing) to LTX-2 Pipelines](/Lightricks/LTX-2/how-to-apply-lora-adapters-ltx-2)

Learn to apply LoRA adapters distilled IC-LoRA and detailing to LTX-2 pipelines Train an adapter save the checkpoint and fuse it into the base model for efficient inference

- Tags: how-to-guide
- Published: 2026-08-20

### [How to Use HDR/EXR Input and Output with LTX-2 Pipelines: Complete Guide](/Lightricks/LTX-2/how-to-use-hdr-exr-input-output-ltx-2)

Learn how to use HDR EXR input and output with LTX-2 pipelines. This guide covers color space flags, EXR data loading, and metadata preservation for seamless HDR conditioning.

- Tags: how-to-guide
- Published: 2026-08-20

### [How FlashAttention, NATTEN, and CuTe DSL Attention Backends Work in LTX-2](/Lightricks/LTX-2/how-do-attention-backends-work-in-ltx-2)

Explore how LTX-2's interchangeable attention backends like FlashAttention, NATTEN, and CuTe DSL optimize performance. Discover automatic selection and fallback mechanisms for seamless GPU integration.

- Tags: internals
- Published: 2026-08-20

### [LTX-2 Quantization Policies and GPU Support: Complete Implementation Guide](/Lightricks/LTX-2/ltx-2-quantization-policies-and-gpu-support)

Explore LTX-2 quantization policies fp8-cast fp8-scaled-mm nvfp4-cast and nvfp4-prequant. Discover GPU support for Ada Lovelace Hopper and BF16-capable CUDA GPUs.

- Tags: how-to-guide
- Published: 2026-08-20

### [Monolith vs Split Checkpoint Layouts in LTX-2: A Complete Guide](/Lightricks/LTX-2/difference-between-monolith-and-split-checkpoint-layouts-ltx-2)

Understand monolith vs split checkpoint layouts in LTX-2. Discover how to choose the best option for your project, balancing ease of use with modular flexibility.

- Tags: deep-dive
- Published: 2026-08-20

### [Custom CUDA Kernels in LTX-2: Complete Build Guide and Kernel Reference](/Lightricks/LTX-2/what-custom-cuda-kernels-does-ltx-2-provide)

Explore LTX-2's custom CUDA kernels, including compiled and JIT-compiled options. Get a complete build guide and reference for efficient GPU acceleration. Learn how to build and integrate them.

- Tags: how-to-guide
- Published: 2026-08-20

### [How to Set Up Multi-GPU Inference with Sequence Parallel and Tiled Data Parallel in LTX-2](/Lightricks/LTX-2/how-to-setup-multi-gpu-inference-ltx-2)

Learn how to set up multi-GPU inference with LTX-2 using Sequence Parallel and Tiled Data Parallel. Optimize transformer inference across multiple GPUs for faster results.

- Tags: how-to-guide
- Published: 2026-08-20

### [Diffusion VAE Decoder Optimization Modes in LTX-2: Complete Guide to `DiffVAEMode`](/Lightricks/LTX-2/what-are-ltx-2-diffusion-vae-decoder-optimization-modes)

Explore the four DiffVAEMode presets in LTX-2: CHUNKED_EAGER, CHUNKED_COMPILE, COMBINED_COMPILE, and BLACKWELL_DSL. Optimize your Diffusion VAE decoder for speed, compilation time, and memory.

- Tags: deep-dive
- Published: 2026-08-20

### [How to Use `torch.compile` for Faster Inference with LTX-2](/Lightricks/LTX-2/how-to-use-torch-compile-for-ltx-2-inference)

Accelerate LTX-2 inference speed using torch.compile. Learn how to set CompilationConfig for reduced overhead and maximum GPU throughput.

- Tags: how-to-guide
- Published: 2026-08-20

### [Memory Optimization Techniques in LTX-2: 9 Methods for Efficient Video Generation](/Lightricks/LTX-2/what-are-ltx-2-memory-optimization-techniques)

Discover 9 memory optimization techniques in LTX-2 to enable efficient video generation on GPUs with limited VRAM. Boost performance and quality.

- Tags: deep-dive
- Published: 2026-08-20

### [How to Choose the Right LTX-2 Pipeline for Your Specific Use Case](/Lightricks/LTX-2/how-to-choose-ltx-2-pipeline-for-use-cases)

Discover how to select the ideal LTX-2 pipeline by evaluating speed quality conditioning and HDR support for your unique video editing audio generation or HDR output needs.

- Tags: how-to-guide
- Published: 2026-08-20

### [How LTX-2's DiT-Based Architecture Works for Audio-Video Generation](/Lightricks/LTX-2/how-does-ltx-2-dit-architecture-work-for-audio-video-generation)

Discover how LTX-2's DiT-based architecture generates synchronized audio and video using a unified latent representation. Explore its innovative approach to media creation.

- Tags: architecture
- Published: 2026-08-20

### [LTX-2 Model Version Detection Rules and Checkpoint Format Differences Explained](/Lightricks/LTX-2/model-version-detection-checkpoint-format-ltx-2-3-vs-ltx-2-5)

Understand LTX-2 model version detection rules and checkpoint format differences between LTX-2.3 and LTX-2.5. Learn how generation is determined and key format variations.

- Tags: deep-dive
- Published: 2026-08-18

### [How LTX‑2's Two‑Stage Pipeline Handles Latent Upsampling Between Stage I and Stage II](/Lightricks/LTX-2/two-stage-pipeline-latent-upsampling-stage-1-stage-2)

Discover how LTX-2's two-stage pipeline manages latent upsampling between stages. Learn about encoder-guided normalization and pixel-shuffle upsampling for efficient refinement.

- Tags: internals
- Published: 2026-08-18

### [How Classifier-Free Guidance (CFG) Works to Blend Conditioned and Unconditioned Predictions in LTX‑2](/Lightricks/LTX-2/classifier-free-guidance-cfg-blend-conditioned-unconditioned-predictions)

Discover how Classifier-Free Guidance (CFG) blends conditioned and unconditioned predictions in LTX-2. Learn how cfg_scale controls the blend to steer diffusion models toward your prompts.

- Tags: deep-dive
- Published: 2026-08-18

### [How to Configure torch.compile for Maximum Inference Performance in LTX-2](/Lightricks/LTX-2/configure-torch-compile-maximum-inference-performance)

Achieve maximum LTX-2 inference performance by configuring torch.compile with specific settings for per-block compilation and CUDA-graph capture. Learn the optimal parameters.

- Tags: performance
- Published: 2026-08-18

### [LTX-2 Pipeline Architecture: How DiffusionStage, VideoDecoder, AudioDecoder, and PromptEncoder Work Together](/Lightricks/LTX-2/ltx-2-pipeline-building-blocks-architecture)

Explore the LTX-2 pipeline architecture. Learn how DiffusionStage, VideoDecoder, AudioDecoder, and PromptEncoder work together for efficient memory usage and stateless reuse. Discover the build-use-free lifecycle.

- Tags: architecture
- Published: 2026-08-18

### [How to Optimize Memory in LTX-2 Using BlockStreamingWrapper and BatchSplitAdapter](/Lightricks/LTX-2/optimize-memory-blockstreamingwrapper-batchsplitadapter)

Optimize LTX-2 memory by streaming transformer weights with BlockStreamingWrapper and splitting batches using BatchSplitAdapter to stay within GPU limits.

- Tags: how-to-guide
- Published: 2026-08-18

### [Euler Ancestral vs Plain Euler vs Res2s Sampler in LTX-2: When to Use Each](/Lightricks/LTX-2/euler-ancestral-vs-euler-res2s-sampler-usage)

Discover when to use Euler Ancestral for stochastic diversity, plain Euler for fast deterministic sampling, and Res2s for high accuracy with fewer steps in LTX-2.

- Tags: deep-dive
- Published: 2026-08-18

### [How the A2Vid Pipeline Leverages Audio Conditioning for Video Generation in LTX-2](/Lightricks/LTX-2/a2vidpipeline-audio-conditioning-video-generation)

Discover how the A2VidPipeline uses audio conditioning to generate synchronized videos. Learn how LTX-2 leverages acoustic cues for cross-modal generation.

- Tags: deep-dive
- Published: 2026-08-18

### [How to Configure Model Offloading (CPU/Disk) to Minimize VRAM Usage in LTX-2](/Lightricks/LTX-2/configure-model-offloading-cpu-disk-reduce-vram-usage)

Minimize VRAM usage in LTX-2 with model offloading. Configure CPU or Disk modes to stream weights and reduce consumption to ~5 GB.

- Tags: how-to-guide
- Published: 2026-08-18

### [How to Integrate LTX-2 with ComfyUI Using the Official Integration](/Lightricks/LTX-2/integrate-ltx-2-with-comfyui-official-integration)

Integrate LTX-2 with ComfyUI easily using the official integration. Install the package and use custom nodes to leverage LTX-2 pipelines for advanced video generation.

- Tags: how-to-guide
- Published: 2026-08-18

### [How DFRPipeline Handles Keyframe Rendering with Spatial Detailing and Temporal Upsampling](/Lightricks/LTX-2/dfrpipeline-keyframe-rendering-spatial-detailing-temporal-upsampling)

Learn how DFRPipeline renders keyframes using spatial detailing and temporal upsampling. Discover its two-stage diffusion process for high-quality video generation.

- Tags: deep-dive
- Published: 2026-08-18

### [How the RetakePipeline Facilitates Selective Regeneration of Specific Time Regions in LTX-2](/Lightricks/LTX-2/retake-pipeline-selective-time-region-video-regeneration)

Learn how the RetakePipeline enables selective video regeneration within specific time regions. Efficiently edit local video sections without reprocessing the entire clip using temporal masks.

- Tags: internals
- Published: 2026-08-18

### [How to Configure HDR Output with EXR Export and Color Spaces (SRGB_LINEAR, ACESCG, ACESCCT) in LTX-2](/Lightricks/LTX-2/hdr-output-configuration-exr-export-color-spaces)

Configure HDR output with EXR export in LTX-2 using SRGB_LINEAR, ACESCG, or ACESCCT color spaces. Learn to generate half-float EXR frames and HLG HEVC masters easily.

- Tags: how-to-guide
- Published: 2026-08-18

### [How Self-Targeted Guidance (STG) Functions in LTX-2 and Choosing Optimal `stg_scale` Values](/Lightricks/LTX-2/self-targeted-guidance-stg-function-optimal-stg-scale)

Discover how Self-Targeted Guidance (STG) functions in LTX-2. Learn to control diffusion denoising and select optimal stg_scale values for better results.

- Tags: deep-dive
- Published: 2026-08-18

### [How to Use the LTX-2 Duration Head for Automatic Frame Count Prediction from Text Prompts](/Lightricks/LTX-2/duration-head-automatic-frame-count-prediction-prompts)

Learn how the LTX-2 Duration Head automatically predicts video frame counts from text prompts. This lightweight model converts text duration into VAE-compliant frame counts.

- Tags: how-to-guide
- Published: 2026-08-18

### [LTX-2 Attention Backends: FlashAttention 3/4, PyTorch SDPA, and NATTEN Explained](/Lightricks/LTX-2/attention-backends-flashattention-pytorch-sdpa-natten-selection)

Explore LTX-2 attention backends like FlashAttention 3/4, PyTorch SDPA, and NATTEN. Learn to select the optimal backend for your hardware and accelerate your models.

- Tags: deep-dive
- Published: 2026-08-18

### [How Multi-GPU Tiled Data Parallel and Sequence Parallel Work in the LTX-2 Framework](/Lightricks/LTX-2/multi-gpu-tiled-data-parallel-sequence-parallel-ltx-2)

Understand how LTX-2 uses multi-GPU Tiled Data Parallel and Sequence Parallel for efficient video generation by distributing spatial tiles and temporal frames across devices.

- Tags: internals
- Published: 2026-08-18

### [LoRA vs. IC-LoRA vs. Distilled LoRA vs. Detailing IC-LoRA in LTX-2: A Complete Technical Comparison](/Lightricks/LTX-2/lora-ic-lora-distilled-lora-detailing-ic-lora-differences)

Understand LoRA, IC-LoRA, Distilled LoRA, and Detailing IC-LoRA in LTX-2. Explore their unique functions from lightweight tuning to pixel-level upsampling for superior image detail.

- Tags: deep-dive
- Published: 2026-08-18

### [How to Configure FP8 Quantization in LTX-2: `fp8-cast` vs `fp8-scaled-mm` for Different GPU Architectures](/Lightricks/LTX-2/fp8-quantization-configuration-fp8-cast-scaled-mm-gpu-architectures)

Configure FP8 quantization in LTX-2 using fp8-cast or fp8-scaled-mm. Optimize performance for different GPU architectures with the --quantization flag or policy objects.

- Tags: deep-dive
- Published: 2026-08-18

### [DiffVAE Modes in LTX-2: CHUNKED_EAGER, NADiffusionDecoder, and ConvDecoder Explained](/Lightricks/LTX-2/diffvae-modes-chunked-eager-nadiffusiondecoder-convdecoder-usage)

Explore LTX-2's DiffVAE modes: CHUNKED_EAGER, NADiffusionDecoder, and ConvDecoder. Understand their trade-offs in compilation, speed, and VRAM for optimal video generation.

- Tags: deep-dive
- Published: 2026-08-18

### [DistilledPipeline vs TI2VidTwoStagesPipeline: Speed and Quality Comparison in LTX-2](/Lightricks/LTX-2/distilledpipeline-vs-ti2vidtwostagespipeline-speed-quality)

Compare DistilledPipeline vs TI2VidTwoStagesPipeline in LTX-2. Discover which pipeline offers faster inference and superior output quality for your video generation needs.

- Tags: performance
- Published: 2026-08-18

### [Debugging Out-of-Memory Errors in LTX-2: GPU Profiling and Gradient-Estimation Sampling](/Lightricks/LTX-2/debugging-out-of-memory-errors-in-ltx-2-with-gradient-estimation-and-profiling)

Debug LTX-2 out-of-memory errors with GPU profiling and gradient-estimation sampling. Reduce VRAM usage and diffusion steps for smoother performance.

- Tags: deep-dive
- Published: 2026-08-15

### [LTX-2 Prompt Enhancement Feature for Improved Generations](/Lightricks/LTX-2/ltx-2-prompt-enhancement-feature-for-improved-generations)

Unlock better generations with the LTX-2 prompt enhancement feature. This pipeline automatically optimizes user prompts for improved video creation using Gemma. Learn how it works.

- Tags: tutorial
- Published: 2026-08-15

### [Integrating LTX-2 with ComfyUI for Workflow Management: A Complete Technical Guide](/Lightricks/LTX-2/integrating-ltx-2-with-comfyui-for-workflow-management)

Learn to integrate LTX-2 with ComfyUI for efficient workflow management. Seamlessly export and load LoRA checkpoints with automatic prefix handling. Get the technical guide now.

- Tags: how-to-guide
- Published: 2026-08-15

### [LTX‑2 STG and Modality Guidance Scales Configuration: A Complete Guide](/Lightricks/LTX-2/ltx-2-stg-and-modality-guidance-scales-configuration)

Master LTX-2 STG and modality guidance scales configuration. Learn to disable STG and control cross-modal isolation for optimal results with this comprehensive guide.

- Tags: how-to-guide
- Published: 2026-08-15

### [Using LTX-2 Spatial and Temporal Upsamplers for Higher Resolution Video Generation](/Lightricks/LTX-2/using-ltx-2-spatial-and-temporal-upsamplers-for-higher-resolution)

Boost video resolution with LTX-2's spatial and temporal upsamplers. Generate higher-quality video by increasing frame size and count. Explore LTX-2 for advanced video generation.

- Tags: tutorial
- Published: 2026-08-15

### [How to Apply Custom LoRAs to the LTX-2 Transformer Model: A Complete Guide](/Lightricks/LTX-2/applying-custom-loras-to-the-ltx-2-transformer-model)

Learn to apply custom LoRAs to the LTX-2 transformer model for efficient fine-tuning. This guide covers PEFT library integration and advanced IC-LoRA features.

- Tags: how-to-guide
- Published: 2026-08-15

### [Understanding the LTX-2 Two-Stage Pipeline Architecture: A Deep Dive into Text-to-Video Generation](/Lightricks/LTX-2/understanding-the-ltx-2-two-stage-pipeline-architecture)

Explore the LTX-2 two-stage pipeline architecture for text-to-video generation. Learn how it creates high-quality videos through low-resolution generation and refinement.

- Tags: deep-dive
- Published: 2026-08-15

### [CPU and Disk Offloading for Large Model Inference in LTX-2](/Lightricks/LTX-2/cpu-and-disk-offloading-for-large-model-inference-in-ltx-2)

Discover how LTX-2 enables large model inference by transparently offloading weights to CPU and disk. Stream weights from RAM or memory-mapped disk to overcome GPU memory limits.

- Tags: performance
- Published: 2026-08-15

### [LTX-2 Attention Backend Options: FlashAttention, PyTorch SDPA, and NADiffusion Decoder Explained](/Lightricks/LTX-2/ltx-2-attention-backend-options-flashattention-pytorch-sdpa-nadiffusiondecoder)

Explore LTX-2 attention backend options including FlashAttention, PyTorch SDPA, and NADiffusion. Discover efficient choices for your models with hardware detection.

- Tags: deep-dive
- Published: 2026-08-15

### [Configuring Generated Keyframes for Enhanced Temporal Resolution in LTX-2](/Lightricks/LTX-2/configuring-generated-keyframes-for-enhanced-temporal-resolution-in-ltx-2)

Enhance LTX-2 video temporal resolution by configuring generated keyframes. Insert more latent frames using --num-generated-keyframes N to shorten diffusion length.

- Tags: deep-dive
- Published: 2026-08-15

### [LTX-2 KeyframeInterpolationPipeline: Generate Smooth Video from Image Keyframes](/Lightricks/LTX-2/ltx-2-keyframeinterpolationpipeline-for-interpolating-between-images)

Generate smooth video from image keyframes with LTX-2 KeyframeInterpolationPipeline. Discover its two-stage diffusion workflow for temporally coherent full-resolution video.

- Tags: tutorial
- Published: 2026-08-15

### [How to Use the LTX-2 DubItPipeline for Re-Voicing and Lip-Sync Synchronization](/Lightricks/LTX-2/ltx-2-dubitpipeline-for-re-voicing-and-lip-sync-synchronization)

Master LTX-2 DubItPipeline for seamless re-voicing and lip-sync synchronization. Preserve speaker identity and facial expressions with this powerful two-stage diffusion process.

- Tags: how-to-guide
- Published: 2026-08-15

### [How to Use the LTX-2 RetakePipeline for Regenerating Specific Video Segments](/Lightricks/LTX-2/using-ltx-2-retakepipeline-for-regenerating-specific-video-segments)

Master LTX-2 RetakePipeline to regenerate specific video segments. Learn how to preserve surrounding frames and refine your videos with this powerful tool.

- Tags: how-to-guide
- Published: 2026-08-15

### [Generating Video from Audio Using LTX-2 A2VidPipelineTwoStage: Complete Implementation Guide](/Lightricks/LTX-2/generating-video-from-audio-using-ltx-2-a2vidpipelinetwostage)

Learn to generate high-resolution videos from audio with LTX-2 A2VidPipelineTwoStage. This guide details the complete implementation of a two-stage diffusion process.

- Tags: how-to-guide
- Published: 2026-08-15

### [LTX-2 DFRPipeline for Detail-Fidelity Rendering with Keyframes: A Complete Guide](/Lightricks/LTX-2/ltx-2-dfrpipeline-for-detail-fidelity-rendering-with-keyframes)

Explore the LTX-2 DFRPipeline for detail-fidelity rendering with keyframes. This guide details its three-stage diffusion process for temporally consistent, high-fidelity video generation.

- Tags: how-to-guide
- Published: 2026-08-15

### [LTX-2 DistilledPipeline vs. Full-Model Pipelines: When to Use Which for Video Generation](/Lightricks/LTX-2/ltx-2-distilledpipeline-vs-full-model-pipelines-when-to-use-which)

Discover when to use LTX-2 DistilledPipeline for faster inference and lower VRAM needs versus full-model pipelines for ultimate quality in video generation.

- Tags: deep-dive
- Published: 2026-08-15

### [Accelerating LTX-2 Transformer Inference with `torch.compile`: A Complete Guide](/Lightricks/LTX-2/accelerating-ltx-2-transformer-inference-with-torch-compile)

Accelerate LTX-2 transformer inference using torch.compile. This guide shows how to compile transformer blocks for faster diffusion models and reduce Python overhead.

- Tags: performance
- Published: 2026-08-15

### [Configuring Block Streaming for Memory‑Constrained GPUs in LTX‑2](/Lightricks/LTX-2/configuring-block-streaming-for-memory-constrained-gpus-in-ltx-2)

Optimize LTX-2 diffusion models for memory-constrained GPUs by configuring block streaming. Reduce VRAM needs by streaming transformer blocks from disk. Learn how to set gpu_slots_count and cpu_slots_count.

- Tags: how-to-guide
- Published: 2026-08-15

### [Automatic Duration Prediction in LTX-2: How to Bypass Manual Frame Counts](/Lightricks/LTX-2/automatic-duration-prediction-in-ltx-2-to-bypass-frame-counts)

Discover automatic duration prediction in LTX-2. Bypass manual frame counts with the DurationHead neural network. Convert seconds to VAE-compatible frames effortlessly.

- Tags: how-to-guide
- Published: 2026-08-15

### [Generating HDR Videos with EXR Output and Color Space Options in LTX-2](/Lightricks/LTX-2/generating-hdr-videos-with-exr-output-and-color-space-options)

Generate HDR videos with EXR output in LTX-2 using SRGB_LINEAR, ACESCG, or ACESCCT color spaces. Leverage HDRICLoraPipeline or the --hdr flag for professional results.

- Tags: how-to-guide
- Published: 2026-08-15

### [Using IC-LoRA for Video-to-Video Transformation with Reference Images in LTX-2](/Lightricks/LTX-2/using-ic-lora-for-video-to-video-transformation-with-reference-images)

Discover how IC-LoRA in LTX-2 transforms videos using reference images. Explore two-stage diffusion for stunning video-to-video generation.

- Tags: tutorial
- Published: 2026-08-15

### [Enabling Multi-GPU Inference with Tensor and Sequence Parallelism in LTX-2](/Lightricks/LTX-2/enabling-multi-gpu-inference-with-tensor-and-sequence-parallelism-in-ltx-2)

Scale video generation with LTX-2 multi-GPU inference. Learn how tensor and sequence parallelism optimize performance using PyTorch distributed primitives in ltx-pipelines.

- Tags: performance
- Published: 2026-08-15

### [How to Use FP8 Quantization for LTX-2 GPU Memory Reduction](/Lightricks/LTX-2/how-to-use-fp8-quantization-for-ltx-2-gpu-memory-reduction)

Reduce LTX-2 GPU memory by 50% using FP8 quantization. Explore FP8-CAST and FP8-SCALED-MM to store model weights in 8-bit format and cut memory usage.

- Tags: performance
- Published: 2026-08-15

### [How to Work with Generated Keyframes in Lightricks LTX-2: A Complete Technical Guide](/Lightricks/LTX-2/work-with-generated-keyframes-feature)

Master LTX-2 generated keyframes for smoother video diffusion. This guide details how to relax temporal compression for higher visual fidelity using CLI or Python.

- Tags: how-to-guide
- Published: 2026-08-14

### [What Are Sigma Schedules and Step Counts per Pipeline in LTX‑2?](/Lightricks/LTX-2/sigma-schedules-step-counts-per-pipeline)

Understand LTX-2 sigma schedules and step counts. Learn how 1D tensors of noise levels guide diffusion and step counts control denoising iterations for optimal results.

- Tags: deep-dive
- Published: 2026-08-14

### [How to Configure the Video VAE Decoder in LTX-2: NADiffusion Decoder vs. Convolutional](/Lightricks/LTX-2/configure-video-vae-decoder-nadiffusiondecoder-convolutional)

Configure LTX-2's video VAE decoder: choose NADiffusionDecoder for quality or convolutional for lower VRAM. Learn control methods like DiffVAEMode and CLI flags.

- Tags: how-to-guide
- Published: 2026-08-14

### [How to Use Prompt Enhancement for Better Results in LTX-2](/Lightricks/LTX-2/use-prompt-enhancement-better-results)

Enhance your LTX-2 video generation with prompt enhancement. Discover how this Gemma-based model improves quality, enriches prompts, and uses reference images for faithful results.

- Tags: how-to-guide
- Published: 2026-08-14

### [How to Integrate LTX-2 with ComfyUI: A Complete Setup Guide](/Lightricks/LTX-2/integrate-ltx-2-with-comfyui)

Integrate LTX-2 with ComfyUI using our complete setup guide. Install the LTX-2 package, clone the ComfyUI-LTXVideo bridge, and place model checkpoints to unlock powerful video generation nodes.

- Tags: how-to-guide
- Published: 2026-08-14

### [How to Debug Attention Patterns and Guidance Issues in LTX-2: A Complete Technical Guide](/Lightricks/LTX-2/debug-attention-patterns-guidance-issues)

Debug LTX-2 attention patterns and guidance issues with this technical guide. Learn to extract keep-masks, isolate attention types, and force backends for accurate results.

- Tags: how-to-guide
- Published: 2026-08-14

### [Distilled vs. Full Model Checkpoints in LTX-2: A Complete Technical Comparison](/Lightricks/LTX-2/differences-distilled-full-model-checkpoints)

Explore LTX-2's distilled vs full model checkpoints. Discover 22B parameter full models for fidelity and 3-4x faster distilled models with optimized inference.

- Tags: deep-dive
- Published: 2026-08-14

### [How to Use A2VidPipeline for Audio-to-Video Generation in LTX-2](/Lightricks/LTX-2/work-with-a2vidpipeline-audio-to-video-generation)

Generate synchronized video from audio using A2VidPipeline in LTX-2. Learn to instantiate A2VidPipelineTwoStage and leverage its two-stage diffusion process for audio-to-video creation.

- Tags: how-to-guide
- Published: 2026-08-14

### [How to Use DurationHead for Automatic Frame Count Prediction in LTX-2](/Lightricks/LTX-2/use-durationhead-automatic-frame-count-prediction)

Learn to use DurationHead for automatic frame count prediction in LTX-2. This lightweight module estimates shot duration without the full diffusion pipeline.

- Tags: how-to-guide
- Published: 2026-08-14

### [How to Optimize Memory with Block Streaming and Batch Splitting in LTX-2](/Lightricks/LTX-2/optimize-memory-block-streaming-batch-splitting)

Optimize LTX-2 memory using block streaming and batch splitting. Run large transformer models on GPUs with 8-10 GB VRAM efficiently.

- Tags: how-to-guide
- Published: 2026-08-14

### [How to Train Custom LoRAs Using ltx-trainer: A Complete Guide for LTX Video Diffusion Models](/Lightricks/LTX-2/train-custom-loras-ltx-trainer)

Learn to train custom LoRAs for LTX video diffusion models. Follow this guide to preprocess data, configure training, and run the ltx-trainer script for powerful AI video generation.

- Tags: how-to-guide
- Published: 2026-08-14

### [How to Reduce Inference Steps with Gradient Estimation in LTX-2 for Faster Video Generation](/Lightricks/LTX-2/reduce-inference-steps-gradient-estimation)

Learn how LTX-2 reduces video generation inference steps by 50% using gradient estimation and velocity correction without sacrificing quality. Optimize your diffusion models today.

- Tags: performance
- Published: 2026-08-14

### [How to Use RetakePipeline for Selective Video Region Regeneration in LTX‑2](/Lightricks/LTX-2/use-retakepipeline-selective-video-region-regeneration)

Learn to regenerate specific video segments with RetakePipeline in LTX-2. Control regeneration time and freeze other parts for precise video editing.

- Tags: how-to-guide
- Published: 2026-08-14

### [How to Implement HDR/EXR Output Workflows with LTX-2: Complete Developer Guide](/Lightricks/LTX-2/implement-hdr-exr-output-workflows)

Unlock HDR/EXR output workflows with LTX-2. This guide details using the CLI flag and HDRICLoraPipeline for advanced HDR transformations and EXR generation. Get your HDR video output today.

- Tags: how-to-guide
- Published: 2026-08-14

### [FlashAttention 3 vs FlashAttention 4 vs PyTorch SDPA: Which LTX-2 Attention Backend Should You Choose?](/Lightricks/LTX-2/select-attention-backend-flashattention-pytorch-sdpa)

Choose the best LTX-2 attention backend. Optimize performance with FlashAttention 4 on CUDA, use FlashAttention 3 for older systems, or PyTorch SDPA for CPU/MPS.

- Tags: comparison
- Published: 2026-08-14

### [How to Configure DFRPipeline Temporal Upsampling Rounds in LTX-2](/Lightricks/LTX-2/configure-dfrpipeline-temporal-upsampling-rounds)

Learn how to configure DFRPipeline temporal upsampling rounds in LTX-2. Set temporal_upsample_rounds to 0 1 or 2 and provide a valid path for smoother results.

- Tags: how-to-guide
- Published: 2026-08-14

### [How to Use the Dub-It Pipeline for Audio Dubbing with Lip Sync in LTX-2](/Lightricks/LTX-2/use-dub-it-pipeline-audio-dubbing-lip-sync)

Master audio dubbing with lip sync in LTX-2 using the Dub-It pipeline. Learn how this innovative process ensures perfect synchronization for your videos.

- Tags: how-to-guide
- Published: 2026-08-14

### [How to Apply IC-LoRA for Video-to-Video Transformations Using LTX-2](/Lightricks/LTX-2/apply-ic-lora-video-to-video-transformations)

Learn to apply IC-LoRA for video-to-video transformations using LTX-2. Train a LoRA with VideoToVideoStrategy and run inference through ICLoraPipeline for stunning results.

- Tags: how-to-guide
- Published: 2026-08-14

### [How to Set Up Model Offloading to CPU or Disk for VRAM Constraints in LTX-2](/Lightricks/LTX-2/setup-model-offloading-cpu-disk-vram-constraints)

Easily manage LTX-2 VRAM constraints with CPU or disk offloading. Learn to use the --offload flag to route model weights and free up GPU memory efficiently.

- Tags: how-to-guide
- Published: 2026-08-14

### [How to Use CFG and STG Guidance Scales for Better Generation Quality in LTX‑2](/Lightricks/LTX-2/use-cfg-stg-guidance-scales-generation-quality)

Master CFG and STG guidance scales in LTX-2 for superior video and audio generation. Tune them with guidance rescale to balance prompt fidelity, temporal consistency, and artifact reduction.

- Tags: tutorial
- Published: 2026-08-14

### [How to Implement FP8 Quantization for Memory Optimization on Hopper GPUs with LTX-2](/Lightricks/LTX-2/implement-fp8-quantization-memory-optimization-hopper-gpu)

Implement FP8 quantization with LTX-2 on Hopper GPUs to cut weight memory by 3× and boost throughput by 1.4×. Reduce memory usage and accelerate your models.

- Tags: how-to-guide
- Published: 2026-08-14

### [DistilledPipeline vs Full-Model Pipelines in LTX-2: Architecture, Performance, and Use Cases Explained](/Lightricks/LTX-2/difference-between-distilledpipeline-and-full-model-pipelines)

Understand DistilledPipeline vs full model pipelines in LTX-2. Learn about their architecture, performance, and use cases for efficient AI model deployment.

- Tags: architecture
- Published: 2026-08-14

### [How to Configure Multi-GPU Inference with LTX-2 Pipelines: A Complete Guide](/Lightricks/LTX-2/how-to-configure-multi-gpu-inference-with-ltx-2-pipelines)

Master multi-GPU inference with LTX-2 pipelines. Learn to subclass MGPURunner, use MGPUController, and stream jobs for efficient SPMD execution across devices. Unlock faster AI.

- Tags: how-to-guide
- Published: 2026-08-14

### [Requirements for FP8 Scaled MM Quantization in LTX-2: A Complete Guide](/Lightricks/LTX-2/what-are-the-requirements-for-using-fp8-scaled-mm-quantization-in-ltx-2)

Learn the requirements for FP8 scaled MM quantization in LTX-2. Use FP8 weights, per-tensor scales, and a PyTorch build with torch.float8_e4m3fn for efficient quantization.

- Tags: how-to-guide
- Published: 2026-06-21

### [How to Use FP8 Cast Quantization via CLI in LTX-2](/Lightricks/LTX-2/how-to-use-fp8-cast-quantization-via-cli-in-ltx-2)

Learn to use FP8 Cast quantization via CLI in LTX-2. Enable on-the-fly FP8 weight storage with bfloat16 inference up-casting by adding --quantization fp8-cast to your LTX-2 pipeline command.

- Tags: how-to-guide
- Published: 2026-06-21

### [FP8 Quantization Backends in LTX-2: FP8-Cast and FP8-Scaled-MM Explained](/Lightricks/LTX-2/what-are-the-supported-fp8-quantization-backends-in-ltx-2)

Explore FP8 quantization backends FP8-Cast and FP8-Scaled-MM in LTX-2. Discover which backend suits your checkpoint format and inference needs. Learn more now.

- Tags: deep-dive
- Published: 2026-06-21

### [How to Disable Guidance for Faster Inference in LTX-2](/Lightricks/LTX-2/how-do-i-disable-guidance-for-faster-inference-in-ltx-2)

Learn how to disable guidance in LTX-2 for faster inference. Set cfg scale to 1.0 and stg scale to 0.0 to roughly halve inference time while keeping base model output.

- Tags: how-to-guide
- Published: 2026-06-21

### [How Modality CFG Improves Audio-Visual Synchronization in LTX-2](/Lightricks/LTX-2/how-can-modality-cfg-improve-audio-visual-synchronization-in-ltx-2)

Discover how Modality CFG enhances audio-visual synchronization in LTX-2. Learn how it amplifies cross-modal guidance to keep video and audio streams perfectly in sync, preventing drift.

- Tags: deep-dive
- Published: 2026-06-21

### [Which Transformer Blocks Are Affected by STG in LTX-2 by Default?](/Lightricks/LTX-2/which-transformer-blocks-are-affected-by-stg-in-ltx-2-by-default)

Discover which transformer blocks STG affects in LTX-2 by default. Learn how STG targets block 29 for video generation.

- Tags: deep-dive
- Published: 2026-06-21

### [How Spatio-Temporal Guidance (STG) Works in LTX-2: A Technical Deep Dive](/Lightricks/LTX-2/how-does-spatio-temporal-guidance-stg-work-in-ltx-2)

Explore Spatio-Temporal Guidance (STG) in LTX-2. Learn how STG enhances video generation by steering the diffusion process for improved temporal coherence. Technical deep dive.

- Tags: deep-dive
- Published: 2026-06-21

### [Default CFG Scales for Video and Audio in LTX-2: Complete Technical Reference](/Lightricks/LTX-2/what-are-the-default-cfg-scales-for-video-and-audio-in-ltx-2)

Discover LTX-2 default CFG scales for video and audio generation. Learn the technical specifics for optimizing your creative projects with Lightricks.

- Tags: api-reference
- Published: 2026-06-21

### [LTX-2 Pipeline for Lip Dubbing and Audio Rephrasing: Implementation Guide](/Lightricks/LTX-2/what-ltx-2-pipeline-is-suitable-for-lip-dubbing-and-rephrasing-audio)

Master lip dubbing and audio rephrasing with the LTX-2 framework. Explore the LipDubPipeline and T2AOneStagePipeline for advanced audio generation and manipulation.

- Tags: how-to-guide
- Published: 2026-06-21

### [How to Generate HDR Video Output with LTX-2](/Lightricks/LTX-2/how-can-i-generate-hdr-video-output-with-ltx-2)

Generate HDR video output using LTX-2's HDRICLoraPipeline. This method applies LogC3 decoding and blends HDR conditioning for linear float tensors saved as EXR. Learn more now.

- Tags: how-to-guide
- Published: 2026-06-21

### [RetakePipeline: The LTX-2 Pipeline for Regenerating Specific Video Regions](/Lightricks/LTX-2/which-ltx-2-pipeline-allows-regenerating-specific-video-regions)

Discover the RetakePipeline in LTX-2 for regenerating specific video regions. Modify select time intervals without affecting surrounding content. Learn more about LTX-2's flexible video editing.

- Tags: deep-dive
- Published: 2026-06-21

### [How to Use LTX-2 for Keyframe Interpolation: Two-Stage Pipeline Guide](/Lightricks/LTX-2/how-do-i-use-ltx-2-for-keyframe-interpolation)

Learn to use LTX-2 for keyframe interpolation. This guide details the two-stage diffusion pipeline for smooth, high-resolution video generation.

- Tags: how-to-guide
- Published: 2026-06-21

### [How to Perform Video-to-Video Transformations with LTX-2: The Complete IC-LoRA Guide](/Lightricks/LTX-2/how-can-i-perform-video-to-video-transformations-using-ltx-2)

Learn video-to-video transformations with LTX-2's IC-LoRA guide. Generate new videos preserving source motion and structure with text-guided changes.

- Tags: how-to-guide
- Published: 2026-06-21

### [Which LTX-2 Pipeline Offers the Fastest Inference Speed?](/Lightricks/LTX-2/which-ltx-2-pipeline-offers-the-fastest-inference-speed)

Discover the fastest inference speed with Lightricks LTX-2. Learn how the DistilledPipeline class achieves optimal performance for your NLP tasks.

- Tags: performance
- Published: 2026-06-21

### [The Quickest LTX-2 Pipeline for Prototyping: One-Stage Text-to-Video Generation](/Lightricks/LTX-2/what-is-the-quickest-ltx-2-pipeline-for-prototyping)

Discover the fastest LTX-2 pipeline for prototyping with the one-stage text-to-video model. Generate videos quickly in a single pass, reducing processing time and memory usage.

- Tags: tutorial
- Published: 2026-06-21

### [Which LTX-2 Pipeline Should You Use for Production Text/Image-to-Video Generation?](/Lightricks/LTX-2/which-ltx-2-pipeline-should-i-use-for-production-textimage-to-video-generation)

Choose TI2VidTwoStagesHQPipeline for production text to video generation. Leverage its two-stage coarse-to-fine strategy and Res2S sampler for superior quality and speed.

- Tags: architecture
- Published: 2026-06-21

### [How Cross-Modality AdaLN Synchronizes Audio and Video Streams in LTX-2](/Lightricks/LTX-2/how-does-cross-modality-adaln-synchronize-audio-and-video-streams-in-ltx-2)

Discover how cross-modality AdaLN synchronizes audio and video in LTX-2. Learn how shared parameters ensure identical temporal conditioning for normalized generative timesteps.

- Tags: internals
- Published: 2026-06-21

### [LTX-2 Transformer Block: Complete Sequence of Operations Explained](/Lightricks/LTX-2/explain-the-sequence-of-operations-within-each-ltx-2-transformer-block)

Uncover the LTX-2 transformer block operations. Explore the sequence of self-attention, cross-attention, and feed-forward layers in this multimodal processing pipeline.

- Tags: deep-dive
- Published: 2026-06-21

### [How LTX-2 Achieves Sub-Frame Alignment Using Cross-Attention: 1-D Temporal RoPE Explained](/Lightricks/LTX-2/how-does-ltx-2-achieve-sub-frame-alignment-using-cross-attention)

LTX-2 achieves sub-frame alignment using 1-D temporal RoPE and cross-attention. This powerful technique synchronizes audio and video at a granularity finer than individual frames.

- Tags: deep-dive
- Published: 2026-06-21

### [Rotary Position Embedding (RoPE) Configurations for Video and Audio Streams in LTX-2](/Lightricks/LTX-2/what-are-the-different-rotary-position-embedding-rope-configurations-for-video-and-audio-streams-in-ltx-2)

Explore Rotary Position Embedding RoPE configurations in LTX-2. Discover SPLIT variant for video 3-D embeddings and 1-D temporal for audio streams, optimizing performance.

- Tags: deep-dive
- Published: 2026-06-21

### [LTX-2 Transformer Blocks: Architecture, Count, and Operations Explained](/Lightricks/LTX-2/how-many-transformer-blocks-are-used-in-ltx-2-and-what-operations-do-they-perform)

Discover the architecture and operations of LTX-2 transformer blocks. Learn how 48 blocks perform self-attention, cross-attention, and joint video-audio generation for advanced AI models.

- Tags: deep-dive
- Published: 2026-06-21

### [How LTX-2 Distributes Model Parameters Between Video and Audio Streams](/Lightricks/LTX-2/what-are-the-parameter-distributions-between-the-video-and-audio-streams-in-ltx-2)

Discover how LTX-2 distributes model parameters for video and audio streams with dedicated layers and bidirectional cross-attention. Learn about its unique approach to multimodal AI.

- Tags: architecture
- Published: 2026-06-21

### [How LTX-2's Asymmetric Dual-Stream Transformer Architecture Processes Video and Audio Jointly](/Lightricks/LTX-2/how-does-ltx-2s-asymmetric-dual-stream-transformer-architecture-function)

Discover how LTX-2's asymmetric dual-stream transformer architecture processes video and audio together. Explore its unique approach to cross-modal interaction for enhanced AI functionality.

- Tags: deep-dive
- Published: 2026-06-21

### [How to Use the LTX-2 Model with ComfyUI: Complete Integration Guide](/Lightricks/LTX-2/ltx-2-comfyui-integration)

Easily integrate the LTX-2 model into ComfyUI with this complete guide. Discover how the LTX-2 checkpoints load seamlessly with the standard Checkpoint Loader for instant use.

- Tags: how-to-guide
- Published: 2026-06-20

### [How to Implement GuidedDenoiser vs SimpleDenoiser for Custom Denoisers in LTX-2](/Lightricks/LTX-2/ltx-2-custom-denoisers-guided-simple)

Learn to implement GuidedDenoiser vs SimpleDenoiser for custom denoisers in LTX-2. Explore differences in CFG, STG, and inference speed to optimize your LTX-2 denoiser.

- Tags: how-to-guide
- Published: 2026-06-20

### [How to Configure Cross-Modal Attention Scaling for Modality Guidance in LTX-2](/Lightricks/LTX-2/ltx-2-modality-guidance-cross-modal-attention)

Configure LTX-2 cross-modal attention scaling for modality guidance. Use av_ca_timestep_scale_multiplier to control secondary modality influence on primary generation.

- Tags: how-to-guide
- Published: 2026-06-20

### [How to Reduce Inference Steps While Maintaining Quality Using Gradient Estimation in LTX-2](/Lightricks/LTX-2/ltx-2-gradient-estimation-reduce-inference-steps)

Reduce LTX-2 inference steps to 20-30 while maintaining quality. Learn how gradient estimation and the specialized Euler sampler accelerate your workflow.

- Tags: performance
- Published: 2026-06-20

### [FlashAttention 4 vs xFormers Configuration for LTX-2 Attention Optimization](/Lightricks/LTX-2/ltx-2-attention-optimization-flashattention-xformers)

Optimize LTX-2 attention with FlashAttention 4 vs xFormers. Discover how LTX-2 automatically selects the best backend for your GPU and needs.

- Tags: performance
- Published: 2026-06-20

### [How to Convert Video Pixels to Latent Space Using LTX-2 Video VAE Encoding](/Lightricks/LTX-2/ltx-2-video-vae-encoding-pixels-to-latent)

Learn how to convert video pixels to latent space with LTX-2 Video VAE encoding. Compress spatial and temporal dimensions efficiently for advanced video processing with this powerful technique.

- Tags: tutorial
- Published: 2026-06-20

### [How Gemma 3 Produces Audio vs Video Embeddings in LTX-2's Text Encoder](/Lightricks/LTX-2/ltx-2-gemma-3-audio-video-embeddings)

Discover how Gemma 3 generates distinct audio and video embeddings within LTX-2's text encoder. Learn about synchronized multimodal generation from single prompts.

- Tags: deep-dive
- Published: 2026-06-20

### [How LTX2Scheduler Controls Sigma Schedules in LTX-2](/Lightricks/LTX-2/ltx2scheduler-control-sigma-schedules)

Discover how LTX2Scheduler controls sigma schedules in LTX-2 by generating adaptive noise-scale sequences with linear shifts and non-linear transformations. Learn the technical details.

- Tags: internals
- Published: 2026-06-20

### [How to Generate HDR Video Output with EXR Export Using LTX-2 HDR IC-LoRA](/Lightricks/LTX-2/ltx-2-hdr-ic-lora-exr-export)

Generate HDR video output by exporting EXR tensors with the LTX-2 HDR IC-LoRA pipeline. Learn how to manually save your HDR tensor for advanced video workflows.

- Tags: how-to-guide
- Published: 2026-06-20

### [How to Use the LTX-2 Retake Pipeline to Regenerate Specific Time Regions](/Lightricks/LTX-2/ltx-2-retake-pipeline-regenerate-time-regions)

Learn to regenerate specific time regions in LTX-2 using the Retake pipeline. Replace video intervals while preserving the rest with TemporalRegionMask and ModalitySpec.

- Tags: how-to-guide
- Published: 2026-06-20

### [How to Use the LipDub Pipeline for Audio-Driven Video Generation in LTX-2](/Lightricks/LTX-2/ltx-2-lipdub-pipeline-audio-driven-video)

Generate lip-synchronized videos with the LTX-2 LipDub pipeline. Learn how to use this audio-driven video generation system for stunning visual results.

- Tags: how-to-guide
- Published: 2026-06-20

### [How the Keyframe Interpolation Pipeline Works in LTX-2: A Two-Stage Diffusion Approach](/Lightricks/LTX-2/ltx-2-keyframe-interpolation-pipeline)

Discover how the LTX-2 keyframe interpolation pipeline uses a two-stage diffusion approach to create smooth, high-quality videos by generating a coarse draft and then refining it.

- Tags: deep-dive
- Published: 2026-06-20

### [How to Configure SingleGPUModelBuilder for Loading LTX-2 Model Checkpoints](/Lightricks/LTX-2/ltx-2-configure-singlegpu-modelbuilder)

Configure SingleGPUModelBuilder to load LTX-2 model checkpoints. Specify configurator, path, and LoRA adapters, then build your PyTorch module for single GPU use.

- Tags: how-to-guide
- Published: 2026-06-20

### [Best Practices for Prompt Engineering in LTX-2: A Complete Technical Guide](/Lightricks/LTX-2/ltx-2-prompt-engineering-best-practices)

Master prompt engineering for LTX-2 with our technical guide. Learn to create structured video prompts for Gemma text encoder, ensuring chronological flow and formatting constraints.

- Tags: best-practices
- Published: 2026-06-20

### [How to Handle Large Batch Sizes with Memory Constraints Using Batch Splitting in LTX-2](/Lightricks/LTX-2/ltx-2-large-batch-size-memory-constraints)

Overcome memory limits with LTX-2 batch splitting. Use BatchSplitAdapter to reduce GPU activation memory for large batches without changing model semantics.

- Tags: how-to-guide
- Published: 2026-06-20

### [How LTX-2's Dual-Stream Transformer Handles Audio-Video Synchronization](/Lightricks/LTX-2/ltx-2-dual-stream-transformer-audio-video-sync)

Discover how LTX-2's dual-stream transformer achieves audio-video synchronization using parallel processing and cross-attention. Learn about frame-level alignment and AdaLN modulation.

- Tags: internals
- Published: 2026-06-20

### [How to Enable and Configure FP8 Quantization in LTX-2 for Lower Memory Usage](/Lightricks/LTX-2/ltx-2-enable-configure-fp8-quantization)

Learn how to enable and configure FP8 quantization in LTX-2 to significantly reduce memory usage and preserve output quality. Save up to 31 GiB on large models with these simple steps.

- Tags: how-to-guide
- Published: 2026-06-20

### [How to Run LTX-2 Inference with Limited VRAM Using Block Streaming: A Complete Guide](/Lightricks/LTX-2/ltx-2-limited-vram-block-streaming)

Run LTX-2 inference with limited VRAM using block streaming. Efficiently process large models on GPUs with only 4GB VRAM by streaming transformer blocks between CPU and GPU.

- Tags: how-to-guide
- Published: 2026-06-20

### [How to Use Image-Conditioned LoRA (IC-LoRA) for Video Transformations in LTX-2](/Lightricks/LTX-2/ltx-2-ic-lora-video-transformations)

Learn to use image-conditioned LoRA IC LoRA in LTX-2 for video transformations. This guide details conditioning generation on reference videos with spatial scaling and temporal subsampling.

- Tags: how-to-guide
- Published: 2026-06-20

### [How 8-Step Distilled Inference Works in LTX-2](/Lightricks/LTX-2/ltx-2-8-step-distilled-inference)

Discover how LTX-2's 8-step distilled inference accelerates diffusion models. Learn about its fixed 8-step schedule and hardcoded sigma tensors for efficient, variable-free configuration.

- Tags: internals
- Published: 2026-06-20

### [How to Use and Fuse Multiple LoRA Adapters in LTX-2](/Lightricks/LTX-2/ltx-2-use-fuse-lora-adapters)

Learn to fuse multiple LoRA adapters in LTX-2 by loading checkpoints, aggregating deltas, and applying updates. Master LoRA fusion with the LTX-2 API.

- Tags: how-to-guide
- Published: 2026-06-20

### [How to Configure CFG and STG Guidance Scales in LTX-2](/Lightricks/LTX-2/ltx-2-configure-cfg-stg-guidance)

Learn to configure CFG and STG guidance scales in LTX-2 to precisely control prompt adherence and temporal consistency. Optimize your results with our quick guide.

- Tags: how-to-guide
- Published: 2026-06-20

### [How the Two-Stage Generation Process Works in LTX-2: A Technical Deep Dive](/Lightricks/LTX-2/ltx-2-two-stage-generation-process)

Explore the LTX-2 two-stage generation process. Discover how it creates low-res video then refines it with distilled LoRA for impressive quality gains and modest inference time.

- Tags: deep-dive
- Published: 2026-06-20