[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-74010":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":16,"subscribersCount":16,"size":16,"stars1d":17,"stars7d":18,"stars30d":19,"stars90d":16,"forks30d":16,"starsTrendScore":20,"compositeScore":21,"rankGlobal":10,"rankLanguage":10,"license":22,"archived":23,"fork":23,"defaultBranch":24,"hasWiki":25,"hasPages":23,"topics":26,"createdAt":10,"pushedAt":10,"updatedAt":30,"readmeContent":31,"aiSummary":32,"trendingCount":16,"starSnapshotCount":16,"syncStatus":33,"lastSyncTime":34,"discoverSource":35},74010,"LTX-2","Lightricks\u002FLTX-2","Lightricks","Official Python inference and LoRA trainer package for the LTX-2 audio–video generative model.","https:\u002F\u002Fltx.io\u002Fmodel\u002Fltx-2",null,"Python",7225,1171,74,81,0,87,218,647,261,40.21,"Other",false,"main",true,[27,28,29],"generative-ai","ltx","ltx-2","2026-06-12 02:03:21","# LTX-2\n\n[![Website](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FWebsite-LTX-181717?logo=google-chrome)](https:\u002F\u002Fltx.io)\n[![Model](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FHuggingFace-Model-orange?logo=huggingface)](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2.3)\n[![Demo](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FDemo-Try%20Now-brightgreen?logo=data:image\u002Fpng;base64,iVBORw0KGgoAAAANSUhEUgAAABQAAAAUCAYAAACNiR0NAAAAAXNSR0IArs4c6QAAAERlWElmTU0AKgAAAAgAAYdpAAQAAAABAAAAGgAAAAAAA6ABAAMAAAABAAEAAKACAAQAAAABAAAAFKADAAQAAAABAAAAFAAAAACy3fD9AAACmElEQVQ4Ea1VP2haYRA\u002FfRo0mESRIIqb2IwxuNUl0CGFQBC6OAWcikMottCpqYtDQIgdQsBFhAjZqiQhbhmySJBOgmNU0EGCg9r61Bivd0ffoykE0iQH37\u002F77n7f3e\u002FuqQFIPB7P\u002FN3d3QeDwfAFEedZ91ghnyH5JM1m87dWq6UavF6vdTKZfDcajW\u002Fp4rE49+wIFMj33Gq1vlNo+kxg758KpiETqP\u002F29vaXweVyqaS0aBfPXEfGFwTjWCwM+KBQoWA4HAJx\u002FKDNvxcmTTGbzYAH8SljOp2C2+2GjY0NqNfrcHFxAXNzc2LDfCuKIq78KBdFOwsgGzidTnA4HHBzcwO9Xg8sFgtsbm7C3t4eVCoVaDQa0O12YXl5GUwmk5z5cZ\u002FPB6PRCNrttgADFQUXFhbw8PAQVVXF3d1dJAeMx+P0zn0Jh8OYz+eRADCRSGAqlcLxeIz7+\u002Fu4tLSEjKUDZrNZ8U4mk0jR4fr6Op6enoru+voa0+k0rq2tYTAYxE6ng9QiSLRgrVZDv9+PFLkA6kUhT+GEC8C8XF5ewtHRkejICShiaDabwPvj42NJm3k7ODiQdDl9Fr0ocqJpdXUVIpEIdz7Y7XZRr6ysQDQahXK5LORvbW1p5rC9vQ2UifAooBqHuVxO0vt72tnZwWq1qqtisRgWCgU5ZzIZPDk50fdUUEmZvxTmAgKBgAxunT\u002FfJpRKJWmhUCgEVDi4uroSG46kWCzC4uKitNVgMICzszOhSgA5fiJZhp4Lbbh1KARpbF65D\u002Flx3vMdP05Vlkf5zKIDyukFJi7N6AVwNAhVsdlsM+LsjaZ56sq8kyQUqs4P6rsAKV49B4x4Padf7Y9Kv9+fEmiBQH8S4Gsa5v8EHpL9VwL7xH8BvwEcd4ccVf02KQAAAABJRU5ErkJggg==)](https:\u002F\u002Fconsole.ltx.video\u002Fplayground)\n[![Paper](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FPaper-PDF-EC1C24?logo=adobeacrobatreader&logoColor=white)](https:\u002F\u002Farxiv.org\u002Fabs\u002F2601.03233)\n[![Discord](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FJoin-Discord-5865F2?logo=discord)](https:\u002F\u002Fdiscord.gg\u002Fltxplatform)\n\n**LTX-2** is the first DiT-based audio-video foundation model that contains all core capabilities of modern video generation in one model: synchronized audio and video, high fidelity, multiple performance modes, production-ready outputs, API access, and open access.\n\n\u003Cdiv align=\"center\">\n  \u003Cvideo src=\"https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F4414adc0-086c-43de-b367-9362eeb20228\" width=\"70%\" poster=\"\"> \u003C\u002Fvideo>\n\u003C\u002Fdiv>\n\n## 🚀 Quick Start\n\n```bash\n# Clone the repository\ngit clone https:\u002F\u002Fgithub.com\u002FLightricks\u002FLTX-2.git\ncd LTX-2\n\n# Set up the environment\nuv sync --frozen\nsource .venv\u002Fbin\u002Factivate\n```\n\n### Required Models\n\nDownload the following models from the [LTX-2.3 HuggingFace repository](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2.3):\n\n**LTX-2.3 Model Checkpoint** (choose and download one of the following)\n  * [`ltx-2.3-22b-dev.safetensors`](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2.3\u002Fblob\u002Fmain\u002Fltx-2.3-22b-dev.safetensors) - [Download](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2.3\u002Fresolve\u002Fmain\u002Fltx-2.3-22b-dev.safetensors)\n  * [`ltx-2.3-22b-distilled-1.1.safetensors`](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2.3\u002Fblob\u002Fmain\u002Fltx-2.3-22b-distilled-1.1.safetensors) - [Download](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2.3\u002Fresolve\u002Fmain\u002Fltx-2.3-22b-distilled-1.1.safetensors)\n\n**Spatial Upscaler** - Required for current two-stage pipeline implementations in this repository\n  * [`ltx-2.3-spatial-upscaler-x2-1.1.safetensors`](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2.3\u002Fblob\u002Fmain\u002Fltx-2.3-spatial-upscaler-x2-1.1.safetensors) - [Download](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2.3\u002Fresolve\u002Fmain\u002Fltx-2.3-spatial-upscaler-x2-1.1.safetensors)\n  * [`ltx-2.3-spatial-upscaler-x1.5-1.0.safetensors`](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2.3\u002Fblob\u002Fmain\u002Fltx-2.3-spatial-upscaler-x1.5-1.0.safetensors) - [Download](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2.3\u002Fresolve\u002Fmain\u002Fltx-2.3-spatial-upscaler-x1.5-1.0.safetensors)\n\n**Temporal Upscaler** - Supported by the model and will be required for future pipeline implementations\n  * [`ltx-2.3-temporal-upscaler-x2-1.0.safetensors`](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2.3\u002Fblob\u002Fmain\u002Fltx-2.3-temporal-upscaler-x2-1.0.safetensors) - [Download](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2.3\u002Fresolve\u002Fmain\u002Fltx-2.3-temporal-upscaler-x2-1.0.safetensors)\n\n**Distilled LoRA** - Required for current two-stage pipeline implementations in this repository (except DistilledPipeline, ICLoraPipeline, and LipDubPipeline)\n  * [`ltx-2.3-22b-distilled-lora-384-1.1.safetensors`](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2.3\u002Fblob\u002Fmain\u002Fltx-2.3-22b-distilled-lora-384-1.1.safetensors) - [Download](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2.3\u002Fresolve\u002Fmain\u002Fltx-2.3-22b-distilled-lora-384-1.1.safetensors)\n\n**Gemma Text Encoder** (download all assets from the repository)\n  * [`Gemma 3`](https:\u002F\u002Fhuggingface.co\u002Fgoogle\u002Fgemma-3-12b-it-qat-q4_0-unquantized\u002Ftree\u002Fmain)\n\n**LoRAs**\n  * [`LTX-2.3-22b-IC-LoRA-Union-Control`](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2.3-22b-IC-LoRA-Union-Control) - [Download](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2.3-22b-IC-LoRA-Union-Control\u002Fresolve\u002Fmain\u002Fltx-2.3-22b-ic-lora-union-control-ref0.5.safetensors)\n  * [`LTX-2.3-22b-IC-LoRA-Motion-Track-Control`](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2.3-22b-IC-LoRA-Motion-Track-Control) - [Download](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2.3-22b-IC-LoRA-Motion-Track-Control\u002Fresolve\u002Fmain\u002Fltx-2.3-22b-ic-lora-motion-track-control-ref0.5.safetensors)\n  * [`LTX-2-19b-IC-LoRA-Detailer`](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2-19b-IC-LoRA-Detailer) - [Download](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2-19b-IC-LoRA-Detailer\u002Fresolve\u002Fmain\u002Fltx-2-19b-ic-lora-detailer.safetensors)\n  * [`LTX-2-19b-IC-LoRA-Pose-Control`](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2-19b-IC-LoRA-Pose-Control) - [Download](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2-19b-IC-LoRA-Pose-Control\u002Fresolve\u002Fmain\u002Fltx-2-19b-ic-lora-pose-control.safetensors)\n  * [`LTX-2-19b-LoRA-Camera-Control-Dolly-In`](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2-19b-LoRA-Camera-Control-Dolly-In) - [Download](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2-19b-LoRA-Camera-Control-Dolly-In\u002Fresolve\u002Fmain\u002Fltx-2-19b-lora-camera-control-dolly-in.safetensors)\n  * [`LTX-2-19b-LoRA-Camera-Control-Dolly-Left`](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2-19b-LoRA-Camera-Control-Dolly-Left) - [Download](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2-19b-LoRA-Camera-Control-Dolly-Left\u002Fresolve\u002Fmain\u002Fltx-2-19b-lora-camera-control-dolly-left.safetensors)\n  * [`LTX-2-19b-LoRA-Camera-Control-Dolly-Out`](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2-19b-LoRA-Camera-Control-Dolly-Out) - [Download](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2-19b-LoRA-Camera-Control-Dolly-Out\u002Fresolve\u002Fmain\u002Fltx-2-19b-lora-camera-control-dolly-out.safetensors)\n  * [`LTX-2-19b-LoRA-Camera-Control-Dolly-Right`](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2-19b-LoRA-Camera-Control-Dolly-Right) - [Download](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2-19b-LoRA-Camera-Control-Dolly-Right\u002Fresolve\u002Fmain\u002Fltx-2-19b-lora-camera-control-dolly-right.safetensors)\n  * [`LTX-2-19b-LoRA-Camera-Control-Jib-Down`](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2-19b-LoRA-Camera-Control-Jib-Down) - [Download](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2-19b-LoRA-Camera-Control-Jib-Down\u002Fresolve\u002Fmain\u002Fltx-2-19b-lora-camera-control-jib-down.safetensors)\n  * [`LTX-2-19b-LoRA-Camera-Control-Jib-Up`](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2-19b-LoRA-Camera-Control-Jib-Up) - [Download](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2-19b-LoRA-Camera-Control-Jib-Up\u002Fresolve\u002Fmain\u002Fltx-2-19b-lora-camera-control-jib-up.safetensors)\n  * [`LTX-2-19b-LoRA-Camera-Control-Static`](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2-19b-LoRA-Camera-Control-Static) - [Download](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2-19b-LoRA-Camera-Control-Static\u002Fresolve\u002Fmain\u002Fltx-2-19b-lora-camera-control-static.safetensors)\n  * [`LTX-2.3-22b-IC-LoRA-HDR`](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2.3-22b-IC-LoRA-HDR) - HDR IC-LoRA and pre-computed text embeddings for `HDRICLoraPipeline`\n  * [`LTX-2.3-22b-IC-LoRA-LipDub`](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2.3-22b-IC-LoRA-LipDub) - [Download](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-2.3-22b-IC-LoRA-LipDub\u002Fresolve\u002Fmain\u002Fltx-2.3-22b-ic-lora-lipdub-0.9.safetensors)\n\n### Available Pipelines\n\n* **[TI2VidTwoStagesPipeline](packages\u002Fltx-pipelines\u002Fsrc\u002Fltx_pipelines\u002Fti2vid_two_stages.py)** - Production-quality text\u002Fimage-to-video with 2x upsampling (recommended)\n* **[TI2VidTwoStagesHQPipeline](packages\u002Fltx-pipelines\u002Fsrc\u002Fltx_pipelines\u002Fti2vid_two_stages_hq.py)** - Same two-stage flow as above but uses the res_2s second-order sampler (fewer steps, better quality)\n* **[TI2VidOneStagePipeline](packages\u002Fltx-pipelines\u002Fsrc\u002Fltx_pipelines\u002Fti2vid_one_stage.py)** - Single-stage generation for quick prototyping\n* **[DistilledPipeline](packages\u002Fltx-pipelines\u002Fsrc\u002Fltx_pipelines\u002Fdistilled.py)** - Fastest inference with 8 predefined sigmas\n* **[ICLoraPipeline](packages\u002Fltx-pipelines\u002Fsrc\u002Fltx_pipelines\u002Fic_lora.py)** - Video-to-video and image-to-video transformations (uses distilled model.)\n* **[KeyframeInterpolationPipeline](packages\u002Fltx-pipelines\u002Fsrc\u002Fltx_pipelines\u002Fkeyframe_interpolation.py)** - Interpolate between keyframe images\n* **[A2VidPipelineTwoStage](packages\u002Fltx-pipelines\u002Fsrc\u002Fltx_pipelines\u002Fa2vid_two_stage.py)** - Audio-to-video generation conditioned on an input audio file\n* **[RetakePipeline](packages\u002Fltx-pipelines\u002Fsrc\u002Fltx_pipelines\u002Fretake.py)** - Regenerate a specific time region of an existing video\n* **[HDRICLoraPipeline](packages\u002Fltx-pipelines\u002Fsrc\u002Fltx_pipelines\u002Fhdr_ic_lora.py)** - Video-to-video with HDR output (linear float frames via LogC3 inverse decode, suitable for EXR export and tonemapping)\n* **[LipDubPipeline](packages\u002Fltx-pipelines\u002Fsrc\u002Fltx_pipelines\u002Flipdub.py)** - Lip dubbing, rephrasing, matching speaker identity (distilled model, single IC-LoRA, Two stages).\n\n### ⚡ Optimization Tips\n\n* **Use DistilledPipeline** - Fastest inference with only 8 predefined sigmas (8 steps stage 1, 4 steps stage 2)\n* **Enable FP8 quantization** - Enables lower memory footprint: `--quantization fp8-cast` (CLI) or `quantization=QuantizationPolicy.fp8_cast()` (Python). Fp8-cast should be used with bf16 checkpoints, it shall downcast them on the fly. For Hopper GPUs with TensorRT-LLM, use `--quantization fp8-scaled-mm` for FP8 scaled matrix multiplication. Fp8-scaled-mm should be used with fp8 checkpoints.\n* **Install attention optimizations** - Use xFormers (`uv sync --extra xformers`) or [Flash Attention 3](https:\u002F\u002Fgithub.com\u002FDao-AILab\u002Fflash-attention) for Hopper GPUs\n* **Use gradient estimation** - Reduce inference steps from 40 to 20-30 while maintaining quality (see [pipeline documentation](packages\u002Fltx-pipelines\u002FREADME.md#denoising-loop-optimization))\n* **Skip memory cleanup** - If you have sufficient VRAM, disable automatic memory cleanup between stages for faster processing\n* **Choose single-stage pipeline** - Use `TI2VidOneStagePipeline` for faster generation when high resolution isn't required\n\n## ✍️ Prompting for LTX-2\n\nWhen writing prompts, focus on detailed, chronological descriptions of actions and scenes. Include specific movements, appearances, camera angles, and environmental details - all in a single flowing paragraph. Start directly with the action, and keep descriptions literal and precise. Think like a cinematographer describing a shot list. Keep within 200 words. For best results, build your prompts using this structure:\n\n- Start with main action in a single sentence\n- Add specific details about movements and gestures\n- Describe character\u002Fobject appearances precisely\n- Include background and environment details\n- Specify camera angles and movements\n- Describe lighting and colors\n- Note any changes or sudden events\n\nFor additional guidance on writing a prompt please refer to \u003Chttps:\u002F\u002Fltx.video\u002Fblog\u002Fhow-to-prompt-for-ltx-2>\n\n### Automatic Prompt Enhancement\n\nLTX-2 pipelines support automatic prompt enhancement via an `enhance_prompt` parameter.\n\n## 🔌 ComfyUI Integration\n\nTo use our model with ComfyUI, please follow the instructions at \u003Chttps:\u002F\u002Fgithub.com\u002FLightricks\u002FComfyUI-LTXVideo\u002F>.\n\n## 📦 Packages\n\nThis repository is organized as a monorepo with three main packages:\n\n* **[ltx-core](packages\u002Fltx-core\u002F)** - Core model implementation, inference stack, and utilities\n* **[ltx-pipelines](packages\u002Fltx-pipelines\u002F)** - High-level pipeline implementations for text-to-video, image-to-video, and other generation modes\n* **[ltx-trainer](packages\u002Fltx-trainer\u002F)** - Training and fine-tuning tools for LoRA, full fine-tuning, and IC-LoRA\n\nEach package has its own README and documentation. See the [Documentation](#-documentation) section below.\n\n## 📚 Documentation\n\nEach package includes comprehensive documentation:\n\n* **[LTX-Core README](packages\u002Fltx-core\u002FREADME.md)** - Core model implementation, inference stack, and utilities\n* **[LTX-Pipelines README](packages\u002Fltx-pipelines\u002FREADME.md)** - High-level pipeline implementations and usage guides\n* **[LTX-Trainer README](packages\u002Fltx-trainer\u002FREADME.md)** - Training and fine-tuning documentation with detailed guides\n","LTX-2 是一个基于 DiT 的音视频生成模型，集成了现代视频生成的所有核心功能，包括同步的音频和视频、高保真度、多种性能模式以及生产就绪的输出。该项目提供了Python推理包和LoRA训练器，支持通过API访问模型，并开放了模型使用权。其核心技术特点在于能够生成高质量的音视频内容，适用于需要创建或编辑多媒体内容的各种场景，如影视制作、广告创意、虚拟现实等。",2,"2026-06-11 03:48:24","high_star"]