Kling 3.0 Omni First-Last-Frame to Video

This is the "pick your style of control" node in Kling's Omni lineup. It takes a start frame as the anchor and then lets you choose how to steer the rest: an end frame for a bookended transition, or up to six additional reference images to nail a character's look, or storyboards for a multi-shot scene. It's the Kling 3.0 Omni model wrapped around every input shape people actually ask for, which makes it the node to reach for when you're not sure which of the specialized Kling nodes fits.

Like the rest of the built-in partner/video category, it's an API node: ships with ComfyUI core, uploads your frames to Comfy Org's backend, calls Kling's omni-video endpoint, and bills your Comfy account in credits. Nothing runs locally.

The inputs that matter

• first_frame - required, the anchor. At least 300×300, aspect between 1:2.5 and 2.5:1.
• end_frame - optional, and here's the sharpest constraint in the node: it cannot be used at the same time as reference_images, and it does not work with storyboards. The validation rejects the combos, but plan your approach before you start wiring.
• reference_images - up to six additional references, and it's this or the end frame, not both. Multiple angles of the same subject is the use case.
• prompt - describes the video; can carry positive and negative descriptions in one box; ignored when storyboards are on.
• duration - 3 to 15 seconds. resolution - 720p/1080p/4K, default 1080p. model_name - kling-v3-omni or kling-video-o1, and o1 pays for its budget pricing with hard limits: 10 seconds max, no 4K, no audio, no storyboards.
• storyboards - the dynamic combo that unfolds into up to 6 segments, each with its own prompt and duration. Segment durations must sum exactly to the global duration - the node validates this and tells you the mismatch.
• generate_audio - off by default, only on kling-v3-omni.

Output is a single VIDEO.

Where people get burned

• The end-frame / reference-images exclusivity. People wire both in expecting "start, end, plus extras" and get a validation error. Decide your control strategy up front: bookends OR references, never both.
• o1 mode's quiet feature cuts. If you're on kling-video-o1, storyboards and audio silently vanish from the feature set and the 10-second cap kicks in. The errors are clear when they fire, but the model pick reads like a simple choice until it isn't.
• Storyboard arithmetic. Same trap as the Omni image-to-video node: segment durations must add up to the global duration. Recoverable, but do the math first.
• Cost compounds. Resolution, duration, model, and audio all multiply the per-second rate. A 4K, 15-second, audio-on generation is the most expensive single thing you can ask this node for - and the badge will happily show you that before you commit.

It's the most flexible Kling node, which means it's also the one with the most ways to misconfigure. Pick your mode deliberately and it's superb.