LTX-2.3 22B Video to Video Trainer
/v1/runLtx23 V2v Trainer by Lightricks - advanced AI model for training. Delivers high-quality results with fast inference, suitable for both creative and production workflows.
Request body
Submit an async generation request. The model field selects the model; other fields are model-specific input parameters.
Model identifier. Set to lightricks/ltx23-v2v-trainer.
Default: lightricks/ltx23-v2v-trainer
The aspect ratio of the generated image.
Allowed values: 21:9, 16:9, 3:2, 4:3, 5:4, 1:1, 4:5, 3:4, 2:3, 9:16
Resolution to use for training. Higher resolutions require more memory.
Allowed values: low, medium, high
Default: medium
The resolution to use for validation.
Allowed values: low, medium, high
Default: high
Target frames per second for validation videos.
Range: 8 to 60
Default: 25
Target frames per second for the video.
Range: 8 to 60
Default: 25
When enabled, the trainer returns a downloadable archive of your preprocessed training data for manual inspection. Use this to verify that your videos, images, and captions were processed correctly before committing to a full training run.
Default: false
If true, videos above a certain duration threshold will be split into scenes.
Default: true
If true, videos will be automatically scaled to the target frame count and fps. This option has no effect on image datasets.
Default: false
The number of frames in validation videos.
Range: 9 to 121
Default: 89
A list of validation inputs with prompts and reference videos.
Default:
URL to zip archive with videos or images. Try to use at least 10 files, although more is better. **Supported video formats:** .mp4, .mov, .avi, .mkv **Supported image formats:** .png, .jpg, .jpeg Note: The dataset must contain ONLY videos OR ONLY images - mixed datasets are not supported. The archive can also contain text files with captions. Each text file should have the same name as the media file it corresponds to.
The aspect ratio to use for validation.
Allowed values: 16:9, 1:1, 9:16
Default: 1:1
A phrase that will trigger the LoRA style. Will be prepended to captions during training.
Default:
The number of training steps.
Range: 100 to 20000
Default: 2000
The duration threshold in seconds. If a video is longer than this, it will be split into scenes.
Range: 1 to 60
Default: 30
Learning rate for optimization. Higher values can lead to faster training but may cause overfitting.
Range: 0.000001 to 1
Default: 0.0002
The rank of the LoRA adaptation. Higher values increase capacity but use more memory.
Allowed values: 8, 16, 32, 64, 128
Default: 32
STG (Spatio-Temporal Guidance) scale. 0.0 disables STG. Recommended value is 1.0.
Range: 0 to 3
Default: 1
Number of frames per training sample. Must satisfy frames % 8 == 1 (e.g., 1, 9, 17, 25, 33, 41, 49, 57, 65, 73, 81, 89, 97).
Range: 9 to 121
Default: 89
A negative prompt to use for validation.
Default: worst quality, inconsistent motion, blurry, jittery, distorted
Probability of conditioning on the first frame during training. Lower values work better for video-to-video transformation.
Range: 0 to 1
Default: 0.1
Response Schema
The submit endpoint returns an accepted generation task. Poll the result endpoint with the returned id for terminal outputs or errors.
Error message if the task failed. Empty on success.
Unique identifier for the generation task.
Model ID used for the prediction.
Array of generated content. Empty when status is not completed.
Status of the task: pending, running, completed, failed, or timeout.
Allowed values: pending, running, completed, failed, timeout
Model capabilities
Capabilities declared by the model registry.
Default: training
Execution mode declared by the model registry.
Default: async