Video workflow nodes
Map prompts and frame references for the selected video generation mode.
All four editor modes call video.generate, with different inputs and compatible models.
| Mode | Input to verify |
|---|---|
Text to Video (text_to_video) | A focused prompt. |
Image to Video (image_to_video) | The starting image in firstFramePath. |
First and last frame video (first_last_frame_to_video) | Opening firstFramePath and ending lastFramePath. |
Reference video (reference_to_video on the palette, character_reference_video in the saved mode) | The reference image in referenceImagePath, or references supported by the selected model. |
Check allowed duration, aspect ratio, resolution, and reference types for the selected model. An explicit reference array takes precedence over the frame slots, including when that array is empty. Changing the mode clears the selected offer in the editor, so reselect a compatible model before running. The references slot accepts an array, but supported media and roles depend on that model.
Start with one shot. After the generation completes, preview the clip and inspect artifactId or artifactPath in the run output before mapping it downstream. If a legacy workflow references videoPath or videoArtifact, confirm those fields are present in its actual run output.