- Category
- AI Tools
- Rank
- No. 987Tools index
- Pricing
- Open Source
- Platform
- cli
- Type
- TOOL
- GitHub
- 9.1k stars
- Latest release
- v1.2.0
- Added
- Aug 17, 2026
About
LTX-2 is an open-weight DiT-based foundation model from Lightricks that generates synchronized audio and video from text prompts in a single model, supporting multiple performance modes (distilled vs. full quality), spatial/temporal latent upscaling, LoRA fine-tuning, and API access via a hosted playground.
What it can do
Generate synchronized audio and video from text prompts
Text prompt → Audio-video file
Switch between distilled and full-quality performance modes
Model configuration selection → Generated video at chosen quality/speed tradeoff
Upscale generated video spatially and temporally
Generated latent video → Higher-resolution/frame-rate video
Fine-tune the model using LoRA
Training data → LoRA fine-tuned model weights
Access model generation via hosted API playground
API request with prompt → Generated audio-video output
Run generation pipelines via command-line interface
CLI commands and parameters → Generated audio-video files
Why it made the leaderboard
Open weights plus API access for synchronized audio-video generation gives builders a self-hostable alternative to closed video models. The performance-mode split matters for anyone budgeting GPU time on generation workloads.
Tags
Tech Stack
Comments (0)
No comments yet
Indexed by a proprietary survey. Corrections welcome.