IsraelAudio + videoUp to 4KOpen weights

Explore LTX-2.3

Generate synchronized video and audio from text, images, video, or audio with Lightricks' open-weight cinematic generation model.

Model

Lightricks/LTX-2.3

Input

Configure your request

Generation mode

Video and audio prompt

Describe the scene, motion, camera, dialogue, sound effects, and ambience.

Negative prompt Optional

Describe visual or audio qualities that should be avoided.

Prompt-enhancement system prompt Optional

Supplying this enables the optional prompt-enhancement processor.

Reference image Optional

Optional starting frame for image-to-video generation.

Reference video Optional

Optional source for video-to-video generation.

Reference audio Optional

Optional audio conditioning input.

Parameters

Generation settings

Width

Must be divisible by 32.

Height

Must be divisible by 32.

Frames per second

Number of frames

Use a value divisible by eight, plus one.

Inference steps

Custom sigmas

Optional custom noise schedule; do not combine with custom timesteps.

Custom timesteps

Optional scheduler timesteps; do not combine with custom sigmas.

Guidance scale

Video STG scale

Video modality scale

Video guidance rescale

Audio guidance scale

Audio STG scale

Audio modality scale

Audio guidance rescale

STG transformer blocks

Block 28 is recommended for LTX-2.3.

Initial noise scale

Number of videos

Decode timestep

Decode noise scale

Cross-timestep guidance

Prompt-enhancement tokens

Prompt-enhancement options (JSON)

Optional generation arguments for the prompt enhancer.

Prompt-enhancement seed

Maximum prompt tokens

Attention options (JSON)

Seed

Output type

Explore LTX-2.3

Configure your request

Response