Kling & Omni
Pick text, still-image, or multimodal reference transport deliberately.
Begin with Kling v3 Text
Kling v3 Text is the text-only preset: it exposes a Text port and is appropriate when no source frame is needed. An image connected elsewhere on the graph does not become an input to this preset.
- In Video Generation select Kling v3 Text; write “A glass marble rolls across a tiled floor, single lateral tracking shot.”
- Choose a supported ratio and inspect the current duration control, then run and watch whether the camera actually tracks laterally.
Animate a single source still
Kling v3 uses the image-to-video route and needs a connected image. The first image also controls framing; a typed prompt alone does not satisfy the image requirement.
- Connect a completed Image or uploaded Reference to the Video node’s Image port, select its @ thumbnail, and choose Kling v3.
- Write only the intended change, such as “steam rises slowly; the cup and camera stay still,” then preview for unwanted shape changes.
Use Omni for a reference-rich image shot
Kling v3 Omni accepts Text, Image, and Video ports but requires a prompt plus at least an image or video reference. On the image route the implementation prepares provider-readable HTTPS images.
- Select Kling v3 Omni, connect the approved appearance images, and choose the intended @ thumbnails.
- Specify which visual detail each image contributes; if the request says the image URL is inaccessible, archive or replace that source before trying again.
Use Omni with motion reference
The current Omni video path submits a prepared video reference through a different endpoint from the image path; the implementation takes at most one video for that route. Do not assume its image-path handling is identical.
- Connect and select a short Video result, then describe which gesture to preserve and what scenery to change.
- If you also connected stills, verify the actual reference selection and result; simplify to a single video when troubleshooting transport errors.
Recheck ports after switching models
The model dropdown changes the preset, input ports, and possible aspect-ratio behavior. A Kling image-to-video source cannot simply be carried unchanged into the text-only route.
- After each model change, inspect the visible left ports and the @ list before pressing Run.
- If an image-required error appears, select Kling v3 Text for a text-only experiment or reconnect an actual image for Kling v3.