LTX 2.5
by Lightricks
Video and audio generated together in one pass, from a shot brief.
What it is
LTX 2.5 Pro "generates synchronized video and audio from a text prompt in a single pass, in a quality-optimized mode for final, high-fidelity output". It is Lightricks' open-source audio-video model, and the synchronisation is the part that matters , sound and picture are decided together rather than matched afterwards.
It reads a prompt like a shot brief. Cinematographic and sensory language does real work here: naming the lens, the light and what should be heard changes the output rather than decorating the request.
It runs at 720p and 1080p, and the example output on its own documentation is 1920x1080 at 25 fps.
What it is good at
Synchronised audio generated with the video, not laid over it. What you hear belongs to what you see.
Detailed cinematographic language is processed rather than skimmed. The camera and lens presets in the Studio were built for models that behave like this.
The Pro mode is explicitly for final, high-fidelity output rather than for iteration.
Where it struggles
- Lightricks publishes no list of limitations for this model, so none is invented here.
- It is built for the final pass, which means it is not the cheapest way to explore. Find the shot on a cheaper model first.
- The rate is per second and rises with resolution. The Studio shows both before you run.
How to prompt it
Write it as a shot, and include the sound. Name the framing, the movement, the light and the finish, then say what should be heard , this is one of the few models where "shallow water running over stones, wind high in the pines" changes the result rather than being ignored.
An example that shows what that means
A wide static shot of a shallow river crossing a pine forest at first light, mist sitting on the water, the camera locked off at knee height. Audio: shallow water running over stones, wind high in the pines, one distant bird.
What we offer here
Straight from the catalogue, so this list never goes stale. The price is the provider’s published rate with our markup on top, and it is on the button before you run anything.
Questions
Does it really generate the audio?
Yes, in the same pass as the video, which is why the two match. Say what should be heard in the prompt and it is treated as part of the brief.
What resolution does it run at?
720p and 1080p. The rate is per second and differs between the two; both are shown before you run.
Is it worth it over a cheaper video model?
For the final pass, often yes , the synchronised audio removes a whole step. For working out what the shot should be, use something cheaper first.