Freedom251 › AI video generators
LTX-Video
by LTX-Video
What LTX-Video is
LTX-Video is an open-source model for generating video from text or images. It supports multiple keyframes, keyframe animation, forward or backward video extension, and video-to-video transformations. The repository describes generation up to native 4K at 50 frames per second with synchronized audio in one pass, while the 13B variant supports videos up to 60 seconds. Available model variants include 2B and 13B versions, distilled and FP8-quantized editions, plus depth, pose, and Canny control models. Users can run inference locally with Python and PyTorch or use API, macOS, self-hosted, and web access. ComfyUI and Diffusers integrations are documented, with hosted links for LTX Studio, Fal.ai, and Replicate. A separate trainer supports full and LoRA fine-tuning. Commercial use is allowed under the Apache-2.0 code license. Guidance recommends resolutions below 720 × 1280, fewer than 257 frames, and prompts under 200 words.
Who it suits
It suits developers, creators, and teams wanting open video generation with local, hosted, or integrated workflows. Fine-tuning support helps users adapting the 2B or 13B variants.
What it does well
- Text-to-video and image-to-video
- Native 4K output described
- Multiple keyframe controls
- Local inference instructions
- Commercial use supported
What to know first
- Guidance recommends below 4K working sizes
- Prompt guidance limits length
- Ongoing development points to LTX-2
Verdict
LTX-Video offers broad generation controls, open code, integrations, and fine-tuning options. Review its usage guidance and project-status note when planning a workflow.
Facts from LTX-Video's own pages
- API access
- Yes · github.com, 19 Sept 2026
- Commercial use
- Yes · github.com, 19 Sept 2026
- Company
- Lightricks says it was founded in 2013. · lightricks.com, 13 Oct 2026
- Conditioning
- It supports image-to-video, multiple keyframes, keyframe animation, forward and backward video extension, and video-to-video transformations. · github.com, 12 Oct 2026
- Control models
- The repository lists depth, pose, and Canny control models for LTX-Video. · github.com, 13 Oct 2026
- Fine-tuning
- Lightricks provides an open-source trainer for full fine-tuning and LoRA fine-tuning of the 2B and 13B variants. · github.com, 13 Oct 2026
- Founded
- 2013 · github.com, 28 Sept 2026
- Headquarters
- Jerusalem, Israel · github.com, 28 Sept 2026
- Hosted access
- The README links to online inference through LTX Studio, Fal.ai, and Replicate. · github.com, 12 Oct 2026
- Image-to-video
- Yes · github.com, 19 Sept 2026
- Input constraint
- For video extension, input segments must have a frame count equal to a multiple of 8 plus 1, and the target frame number should be a multiple of 8. · github.com, 12 Oct 2026
- Integrations
- The repository documents ComfyUI and Diffusers integrations and recommends its ComfyUI workflow for best results. · github.com, 13 Oct 2026
- License
- The GitHub repository identifies its code license as Apache-2.0. · github.com, 13 Oct 2026
- Local requirements
- The codebase was tested with Python 3.10.5 and CUDA 12.2; on macOS, MPS was tested with PyTorch 2.3.0. · github.com, 13 Oct 2026
- Local use
- The project provides instructions for running inference locally with Python and PyTorch. · github.com, 12 Oct 2026
- Long generations
- The repository says LTX-Video 13B supports videos up to 60 seconds. · github.com, 13 Oct 2026
- macOS support
- The README says MPS was tested on macOS with PyTorch 2.3.0 and should support PyTorch 2.3 or 2.6 and later. · github.com, 12 Oct 2026
- Maximum resolution
- 4k · github.com, 19 Sept 2026
- Maximum video length
- 1 · github.com, 19 Sept 2026
- Models
- The repository lists 2B and 13B model variants, including distilled and FP8 quantized versions. · github.com, 12 Oct 2026
- Output
- The README describes generation up to 50 FPS at native 4K resolution with synchronized audio in one pass. · github.com, 12 Oct 2026
- Project status
- The repository says LTX-2 is now the primary home for LTX development. · github.com, 13 Oct 2026
- Purpose
- LTX-Video is an open video-generation model that supports text-to-video and image-to-video generation. · github.com, 12 Oct 2026
- Resolution and frame rate
- The repository says the model can generate videos up to 50 FPS at native 4K resolution with synchronized audio in one pass. · github.com, 13 Oct 2026
Where LTX-Video stands on Freedom251
- AI video generators#72 of 242