LTX 2.3 22B

POST/v1/run

Ltx 2.3 22b Reference Video To Video Lora is Lightricks's video-to-video AI model. Transform, enhance, and edit video content using text prompts - from style changes to object manipulation and scene modification.

Request body

Submit an async generation request. The model field selects the model; other fields are model-specific input parameters.

stringmodelrequired

Model identifier. Set to lightricks/ltx-2.3-22b/reference-video-to-video/lora.

Default: lightricks/ltx-2.3-22b/reference-video-to-video/lora

stringpromptrequired

The prompt to generate the video from.

stringimagerequired

An optional URL of an image to use as the first frame of the video.

Optional<string>end_image

The URL of the image to use as the end of the video.

Optional<string>video

The URL of the video to reference.

Optional<string>audio

An optional URL of an audio to use as the audio for the video. If not provided, any audio present in the input video will be used.

Optional<boolean>generate_audio

Whether to generate audio for the video.

Default: true

Optional<integer>num_inference_steps

The number of inference steps to use.

Range: 8 to 50

Default: 40

Optional<number>audio_strength

Audio conditioning strength. Lower values represent more freedom given to the model to change the audio content.

Range: 0 to 1

Default: 1

Optional<number>video_cfg_scale

The Classifier-Free Guidance (CFG) scale for the video. Higher values result in more consistent and focused video content.

Range: 1 to 20

Default: 3

Optional<number>video_strength

Video conditioning strength. Lower values represent more freedom given to the model to change the video content.

Range: 0 to 1

Default: 1

Optional<integer>num_frames

The number of frames to generate.

Range: 9 to 481

Default: 121

Optional<string>scheduler

The scheduler to use.

Allowed values: ltx2, linear_quadratic, beta

Default: ltx2

Optional<array<object>>loras

The LoRAs to use for the generation.

Optional<number>distill_lora_first_pass_scale

The scale of the distill LoRA to use for the first pass. Set to 0 to disable.

Range: 0 to 1

Default: 0.2

Optional<boolean>match_input_fps

When true, match the output FPS to the input video's FPS instead of using the default target FPS.

Default: true

Optional<number>audio_rescaling_scale

The rescaling scale for the audio. Controls the ratio between classifier-free guidance and spatiotemporal guidance.

Range: 0 to 1

Default: 0.7

Optional<string>preprocessor

The preprocessor to use for the generation.

Allowed values: depth, canny, pose, none

Default: none

Optional<number>audio_stg_scale

The Spatiotemporal Guidance (STG) scale for the audio. Higher values result in more consistent and focused audio content.

Range: 0 to 20

Default: 0

Optional<number>distill_lora_second_pass_scale

The scale of the distill LoRA to use for the second and subsequent passes.

Range: 0 to 1

Default: 0.5

Optional<string>video_size

The size of the generated video.

Allowed values: auto, square_hd, square, portrait_4_3, portrait_16_9, landscape_4_3, landscape_16_9

Default: auto

Optional<number>audio_cfg_scale

The Classifier-Free Guidance (CFG) scale for the audio. Higher values result in more consistent and focused audio content.

Range: 1 to 20

Default: 7

Optional<number>gradient_estimation_gamma

The gamma of gradient estimation during denoising. Set to 0 to disable.

Range: 0 to 10

Default: 2

Optional<number>fps

The frames per second of the generated video.

Range: 1 to 60

Default: 24

Optional<boolean>match_video_length

When enabled, the number of frames will be calculated based on the video duration and FPS. When disabled, use the specified num_frames.

Default: true

Optional<string>video_quality

The quality of the generated video.

Allowed values: low, medium, high, maximum

Default: high

Optional<string>ic_lora_type

The IC LoRA type to use for the generation.

Allowed values: match_preprocessor, union, detailer, none

Default: union

Optional<string>camera_lora

The camera LoRA to use. This allows you to control the camera movement of the generated video more accurately than just prompting the model to move the camera.

Allowed values: dolly_in, dolly_out, dolly_left, dolly_right, jib_up, jib_down, static, none

Default: none

Optional<number>audio_modality_scale

The modality scale for the audio. Controls the ratio between video and audio modalities.

Range: 0 to 10

Default: 3

Optional<number>video_modality_scale

The modality scale for the video. Controls the ratio between video and audio modalities.

Range: 0 to 10

Default: 3

Optional<number>video_rescaling_scale

The rescaling scale for the video. Controls the ratio between classifier-free guidance and spatiotemporal guidance.

Range: 0 to 1

Default: 0.7

Optional<boolean>use_restart_sampling

Whether to use restart sampling. This will inject a small amount of noise during each denoising step, which can help improve the quality of the generated video.

Default: false

Optional<number>video_stg_scale

The Spatiotemporal Guidance (STG) scale for the video. Higher values result in more consistent and focused video content.

Range: 0 to 20

Default: 0

Optional<string>video_write_mode

The write mode of the generated video.

Allowed values: fast, balanced, small

Default: balanced

Optional<boolean>use_multiscale

Whether to use multi-scale generation. If True, the model will generate the video at a smaller scale first, then use the smaller video to guide the generation of a video at or above your requested size. This results in better coherence and details.

Default: true

Optional<number>camera_lora_scale

The scale of the camera LoRA to use. This allows you to control the camera movement of the generated video more accurately than just prompting the model to move the camera.

Range: 0 to 1

Default: 1

Optional<string>video_output_type

The output type of the generated video.

Allowed values: X264 (.mp4), VP9 (.webm), PRORES4444 (.mov), GIF (.gif)

Default: X264 (.mp4)

Optional<integer>seed

The seed for the random number generator.

Response Schema

The submit endpoint returns an accepted generation task. Poll the result endpoint with the returned id for terminal outputs or errors.

Optional<string>error

Error message if the task failed. Empty on success.

stringidrequired

Unique identifier for the generation task.

Optional<string>model

Model ID used for the prediction.

Optional<array>outputs

Array of generated content. Empty when status is not completed.

stringstatusrequired

Status of the task: pending, running, completed, failed, or timeout.

Allowed values: pending, running, completed, failed, timeout

Model capabilities

array<string>capability_tagsrequired

Capabilities declared by the model registry.

Default: video-to-video

stringexecution_moderequired

Execution mode declared by the model registry.

Default: async

# 1. Submit a generation request
curl -X POST https://api.sandbase.ai/v1/run \
  -H "Authorization: Bearer $SANDBASE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "lightricks/ltx-2.3-22b/reference-video-to-video/lora",
  "prompt": "style: cinematic/realistic. A man wearing a baseball cap walks through a modern town. He holds a coffee cup.",
  "image": "https://static.sandbase.ai/examples/input.png",
  "video": "https://static.sandbase.ai/examples/lightricks/ltx-2.3-22b/reference-video-to-video/lora/output_url_0.mp4",
  "audio": "https://static.sandbase.ai/examples/input.mp3"
}'

# 2. If the response is still running, poll the returned id
curl https://api.sandbase.ai/v1/run/run_abc123 \
  -H "Authorization: Bearer $SANDBASE_API_KEY"

Response202 Accepted

{
  "id": "run_abc123",
  "model": "lightricks/ltx-2.3-22b/reference-video-to-video/lora",
  "status": "running",
  "outputs": []
}