LTX-2.3
A second take on text-to-video and image-to-video, up to 2K, with audio and a negative prompt.
- Powers
- Create Video, Animate Image
- Modes
- Text to video · Image to video
- GPU class
- 80 GB Ada Pro or 96 GB Blackwell
- In a fresh install
- Off; one switch in Settings → Models





01
What it is
LTX-2.3 is Lightricks' open-weight video model, run here as the 22B dev checkpoint with its distilled LoRA on top. It is the alternative to MiniMax H3 in Create Video and Animate Image: a different look, resolution presets from SD to 2K, and a negative prompt, which MiniMax does not have.
Clips default to eight seconds at 30 fps. Above Full HD the app drops the frame rate to 30 and the length to ten seconds to keep the job inside the worker's memory.
02
Tools it powers
- Create VideooptionA clip from a text prompt, with length, aspect and audio controls.Also: MiniMax H3
- Animate ImageoptionA still becomes a clip; the picture sets the composition.Also: MiniMax H3
03
Modes
- Text to videoPrompt in, clip out, with audio. Quality presets SD, HD, Full HD and 2K, horizontal or vertical.
- Image to videoOne input image sets the composition.
04
Settings exposed
- QualitySD (640 x 480), HD (1280 x 736), Full HD (1920 x 1088) or 2K (2048 x 1152), horizontal or vertical.
- Length and frame rateDefault 8 s at 30 fps; 2K is capped at 10 s and 30 fps.
- AudioGenerated with the picture; can be switched off.
- Negative promptWhat to keep out of the shot.
- Steps1 to 30, default 8.
- SeedRandom, or fixed.
- Custom LoRAsNot supported on this model.
Other video models
- MiniMax H3Create Video (default), Animate Image (default), First & Last Frame, Reference to Video, Multiframe Reference
- SCAIL-2 Motion ControlMotion Control (default)
- WAN 2.2 AnimateMotion Control