LTX-2.5 Generates 10-Second 720p Video in 6.8 Seconds With Native ComfyUI Integration
Core Highlights
The field of video generation has seen another major speed breakthrough. LTX has launched its new-generation model LTX-2.5, which features native integration with the ComfyUI workflow so that creators can call it without complex configuration. The most eye-catching feature is its extremely fast generation speed: under a hardware configuration of two Nvidia GB200 cards, producing a ten-second, 720p-resolution video takes only 6.8 seconds. Simply put, it is almost instant from input to output, pushing the latency of video generation to an unprecedented low level. This performance reframes what users can expect from a local or self-hosted video pipeline and makes real-time iteration feel practical for the first time. For teams that previously waited minutes for a single clip, the change is not incremental but qualitative, and it changes how creators think about drafting and revision. The milestone matters because speed, not only visual quality, is becoming a primary axis of competition in generative video, and LTX is betting that throughput will win daily usage. If a creator can generate, watch, and regenerate a clip within the same train of thought, the creative process becomes a conversation with the model rather than a slow and disconnected queue.
Specific Capabilities and Event Timeline
The LTX-2.5 released this time centers on native ComfyUI integration, which means node-based workflows can drive the model directly and greatly lower the barrier to use. The accompanying LTX-2.5 Fast version focuses on cost and audio capability, and it can directly generate 720p video with a sound track. According to IT之家, the combined cost of a ten-second clip is about 0.90 US dollars, billed at 0.09 US dollars per second. For small and mid-sized teams, this price is quite attractive because it makes short-form video production budgetable at scale. The free tier for smaller organizations further lowers the risk of experimenting with the model in production. Being able to hear the result immediately, rather than syncing audio in a separate step, also shortens the creative loop and removes a common post-production bottleneck. The combination of speed, built-in audio, and low price is what makes the release notable rather than just another routine model drop in an increasingly crowded field of generative video tools.
Technical Details
Behind the speed is LTX's continued investment in model architecture and inference optimization. Achieving a generation time of 6.8 seconds under a two-GB200 configuration indicates deep tuning in parallel computing and memory utilization. Native ComfyUI integration also means the model can be embedded into existing production pipelines as a node, which is convenient for batch and controllable generation. However, the official side has not yet disclosed more details about the model's parameter count and training data. What is visible is the emphasis on throughput and integration rather than on headline benchmark numbers, suggesting a product mindset oriented toward practitioners. The engineering choices clearly favor predictable, repeatable runs inside a studio's existing tooling, and that practical orientation is likely to appeal more to working teams than to benchmark watchers who care only about leaderboard positions and abstract quality scores.
Comparison with Competitors
Horizontally, mainstream video models such as Runway, Kling, and Sora mostly compete on image quality and duration, while LTX-2.5 chooses speed and low cost as its differentiating selling points. Simply put, others compete on a cinematic feel, while it competes on instant usability inside a workflow. For short-video and advertising production scenarios that require a lot of trial and error and quick drafts, this low-latency approach has clear advantages. It is not trying to be the most beautiful model on the leaderboard; it is trying to be the one you can actually use dozens of times per hour without blowing up your budget. That positioning opens a segment the incumbents have largely left alone, and it gives smaller creators a tool that fits their own cadence rather than the slower, more deliberate cadence of a traditional film studio.
Industry Impact and Use Cases
LTX-2.5 is freely available to organizations with annual recurring revenue below 10 million US dollars, a strategy clearly intended to rapidly expand the user base and capture the minds of creators. For independent developers, small studios, and e-commerce teams, this is a low-cost entry point to high-quality video generation. In the long run, native ComfyUI integration will push it into automated production pipelines, becoming a link in the video assembly line rather than an isolated web tool. As more creators wire it into their daily routines, LTX-2.5 could become the default engine for fast, iterative, audio-aware video drafts across a wide range of content businesses. The free threshold also means many startups will qualify on day one, accelerating adoption and giving LTX a broad top-of-funnel that can later convert into paid usage as those teams grow and their video needs scale up.
Who Should Use It and Caveats
LTX-2.5 is at its best for independent creators, small studios, e-commerce teams, and short-video advertisers, especially those who need to iterate quickly and produce many rough drafts before settling on a final cut. Organizations with under 10 million dollars in annual recurring revenue can use it for free, which keeps the barrier to entry very low and invites a wide pool of experimenters. A caveat is that the headline 6.8-second speed depends on top-tier hardware such as two NVIDIA GB200 accelerators, so ordinary consumer GPUs will not deliver anything close to that pace, and buyers should calibrate expectations against their own machines. The model also leans into speed and low cost rather than cinematic polish, so it still trails Sora and Kling on image fidelity and film-like texture. If what you need is a finished, carefully retouched piece rather than a quick storyboard, LTX-2.5 is better treated as a draft tool than as the final delivery pipeline.