Filmmakers
Plan connected sequences with camera cuts and consistent scene direction.
LTX 2.5 generates video with synchronized sound from text or images. Native multi-shot generation connects scenes while maintaining the subject, lighting and voice, with Fast and Pro variants for different creative priorities.
Explore six published scenes, then apply an editable starting prompt to create your own version. Results vary with your model and settings.
LTX 2.5
Desert rider across five shots
LTX 2.5
Rooftop running sequence
LTX 2.5
Canyon flight into the sky
LTX 2.5
Molten bronze and steam
LTX 2.5
Running shoe in the rain
LTX 2.5
Ice sculpture in blue light
Lightricks’ open-weights model for connected video and sound.
Lightricks released LTX 2.5 on August 11, 2026. It combines visual generation with audio and introduces native multi-shot scenes, a new video decoder and stronger interpretation of creative instructions.
Fast prioritizes efficient iteration, while Pro prioritizes visual fidelity. Official model documentation lists output up to 4K, with duration limits that depend on variant, resolution and frame rate. Open weights also support local deployment and fine-tuning under the model’s license.
Start with a scene description or visual material and shape the result:
Six capabilities for planning shots, maintaining continuity and shaping a video.
Describe connected shots in one generation, with continuity in characters, environments, lighting and voices.
Create motion, dialogue and environmental audio together to give a scene a coherent soundscape.
Begin with a written scene or an opening image, then specify the action and camera direction.
Use Fast for speed-focused iteration or Pro when finer visual detail matters more.
Official options reach 4K and 50 fps. Fast clips above 1080p or at 48/50 fps are limited to 10 seconds; Pro clips are also limited to 10 seconds.
The official model can choose a clip length from the requested action, within its variant and output limits. A fixed ending frame requires a fixed duration.
Develop visual stories and test different ways to present an idea.
Plan connected sequences with camera cuts and consistent scene direction.
Explore product motion, lighting and material detail in short visual studies.
Illustrate a short sequence of actions with a focused visual explanation and sound.
Develop landscape and portrait video concepts around a shared subject or message.
Move from a scene idea to a downloadable clip in three steps.
Write the subject, setting and action. Add camera direction and sound, or apply an example and adapt its starting prompt.
Select the model, aspect ratio, duration, resolution and audio option shown in the generator. Review the displayed credit cost before generating.
Review the resulting motion and sound. Adjust your brief for another version, then download the clip you want to use.
Official model capabilities; available controls follow the model selected in the generator.
| Developer | Lightricks |
| Released | August 11, 2026 |
| Variants | Fast and Pro |
| Starting inputs | Text or image; audio-driven generation is also documented |
| Official output resolutions | 720p, 1080p, 1440p and 4K |
| Frame rates | 24, 25, 48 or 50 fps |
| Fast duration | 6–20 seconds in 2-second steps at 720p/1080p and 24/25 fps; otherwise 6, 8 or 10 seconds |
| Pro duration | 6, 8 or 10 seconds |
| Aspect ratios | 16:9, 9:16, 1:1, 3:2 and 2:3 |
| Automatic duration | Text-to-video and image-to-video; cannot combine with an ending frame |
Last updated: November 20, 2026
Compare published model specifications using the same variant and output conditions.
| Specification | LTX 2.5 | LTX 2.3 |
|---|---|---|
| Variants | Fast and Pro | Fast and Pro |
| Official maximum resolution | 4K | 4K |
| Frame rates | 24, 25, 48 or 50 fps | 24, 25, 48 or 50 fps |
| Fast at 720p/1080p, 24/25 fps | 6–20 seconds in 2-second steps | 6–20 seconds in 2-second steps |
| Pro duration | 6, 8 or 10 seconds | 6, 8 or 10 seconds |
| Documented aspect ratios | 16:9, 9:16, 1:1, 3:2, 2:3 | 16:9, 9:16 |
| Native multi-shot and automatic duration | Documented | Not listed on the model specification page |
LTX 2.5 Advantages: Adds documented native multi-shot generation and automatic duration, with square and additional landscape and portrait formats.
LTX 2.3 Advantages: Provides an earlier Fast/Pro workflow with the same published maximum resolution and core frame-rate options.
Choose LTX 2.5 when connected shots or its newer composition options matter. LTX 2.3 can suit an existing workflow built around its documented portrait or landscape formats.
Choose a plan for your projects. Credits depend on the actual model and settings selected in the generator; review the displayed total before creating.
Answers about inputs, connected shots, output limits and credits.
Explore the LTX 2.5 AI Video Generator page on Picir AI. Browse examples, adapt a starting prompt and choose your model and settings in the generator.
Lightricks released LTX 2.5 on August 11, 2026. It is an open-weights video and audio model with Fast and Pro variants.
It creates multiple connected shots in a single generation. Repeating key subject, clothing and setting details in your brief helps communicate the continuity you want.
Fast focuses on speed and supports clips up to 20 seconds under specific output settings. Pro focuses on higher visual fidelity, with 6, 8 or 10-second clips in the official support matrix.
The official model supports 4K, but Fast clips at 1440p or 4K are limited to 10 seconds. The 20-second maximum applies to Fast at 720p or 1080p and 24 or 25 fps. Pro clips are limited to 10 seconds.
The LTX model family documents text, image and audio-driven creation. An image can establish the opening composition, while an audio-driven workflow follows the supplied soundtrack. Use the input modes offered by your selected generator model.
Credits depend on the actual selected model and its settings, including duration and resolution. Review the total displayed in the generator and choose a plan that fits your projects.
The button fills an editable creative starting prompt, without starting a generation. The examples illustrate published scenes; results depend on your selected model and settings. Review the output before sharing and only use reference material you have permission to use.