New LTXV 0.9.8 Model Live on fal
We are excited to announce a major update to LTXV (LTX-Video) by Lightricks, now the first gen-Video model capable of long-form generation, with high controllability out-ranking all open source models in lightning-fast speed. This breakthrough introduces a new class of creative storytelling tools for developers and creators.
LTXV 0.9.8: Redefining What Video Models Can Do
The latest LTXV models on fal combine high-resolution output, advanced LoRA-based control, and up to 30x faster inference than any comparable model, all without sacrificing visual fidelity.
Try the updated LTXV models on fal:
With LTXV, you can:
- Apply pose, depth, and canny LoRAs throughout long video sequences
- Upscale to crisp 1080p and generate consistent, artifact-free motion with precision
Whether you’re starting from a single image or key frames, LTXV gives you unmatched flexibility across any platform or pipeline. Multi-scale rendering and cutting-edge deployment mean you get fast, high-quality results every time.
All of this is fully integrated with fal’s infrastructure, making it easy to experiment, prototype, and deploy long-form AI video inside your workflows.
What's New in This Release
This update introduces several key enhancements that unlock new capabilities for real-time AI video:
- Create beyond 30 seconds: LTXV breaks the 30-second generation limit with an 8x increase over the industry standard. The model generates live video as it renders, enabling coherent long-form storytelling for use cases like ads, game cutscenes, or educational explainers.
- Autoregressive sequence conditioning: Each frame block is conditioned on the last, allowing motion, context, and narrative to evolve naturally. Think of it like writing a story one sentence at a time - except it's video, in real time.
- Continuous Control LoRAs: Apply pose, depth, and canny inputs not only at the start, but throughout the full clip. This allows dynamic motion capture, persistent framing, or stylistic control across long sequences.
In addition to the new long-form capabilities, LTXV still supports:
- High-Quality Output: Utilizes advanced multiscale rendering technology to generate sharp, smooth videos with fewer artifacts and improved Prompt Understanding
- Full Creative Control: Key Frame Conditioning for structured video, character and scene-level motion adjustment and spatiotemporal guidance for frame-level editing.
- Fast Performance: Generate videos 30 times faster than comparable models, enabling real-time creation and rapid iteration.
- Built for Efficiency: With 13 billion parameters and built upon a DiT-based architecture, LTXV delivers exceptional video generation capabilities.
- Versatile Input Support: Supports multiple input types, text-to-video & image-to-video, allowing for diverse video creation workflows.
Real-World Use Cases
LTXV’s new architecture opens the door to a wide-range of business applications, built to scale:
- Scalable Exploration and creative thinking at scale
- Live-generated ads and social content from prompt to publish
- Real-time cutscenes rendered from gameplay input
- Stage-synced AR characters reacting to live performers
- Adaptive explainer videos that respond to learner feedback in real time
Start Creating Today
Ready to build with the future of AI video? Access LTXV through fal’s API and start experimenting with real-time, long-form generation. Each prompt is a launchpad for creative control, coherence, and next-level storytelling.