Kling 3.0 Omni Image to Video

This is Kling's flagship reference node, and it's the one that most directly answers the question "how do I keep the same character across a whole generated scene?" It takes up to seven reference images of a subject, plus a prompt, and generates a video starring that subject - with the twist that the whole generation is one Kling 3.0 Omni model call, so the consistency is baked in rather than stitched together. You're not chaining five image-to-video clips and hoping the face holds; you're giving the model a stack of references and letting it hold the face.

Like every node in the built-in partner/video category, it's an API node: ships with ComfyUI core, uploads your images to Comfy Org's backend, calls Kling's omni-video endpoint, and bills your Comfy account in credits. The storyboards option and generate_audio are the extra headroom that justify the "Omni" name.

The inputs that matter

• reference_images - up to seven. This is the consistency engine. Multiple angles of the same person or subject is exactly what the model wants; each must be at least 300×300 with an aspect ratio between 1:2.5 and 2.5:1.
• prompt - describes the video content, and the tooltip notes it can carry both positive and negative descriptions in one box. It's ignored when storyboards are on.
• model_name - kling-v3-omni (the full experience) or kling-video-o1 (Kling's budget reasoning-tier model). Read the fine print below - o1 gives up a lot.
• duration - 3 to 15 seconds. aspect_ratio - 16:9, 9:16, 1:1. resolution - 720p/1080p/4K, default 1080p.
• storyboards - a dynamic combo that unfolds into 1–6 segments, each with its own prompt and duration, letting one generation produce a multi-shot mini-sequence. Here's the catch that trips everyone: the segment durations must add up exactly to the global duration. It's validated, so you'll get a clear error, but do the arithmetic before you render.
• generate_audio - Kling 3.0 Omni can produce audio; it's off by default and only available on kling-v3-omni.

Output is a single VIDEO.

Where people get burned

• kling-video-o1 is a feature desert. It caps at 10 seconds, refuses 4K, and rejects both audio and storyboards. If you pick it expecting the full Omni experience, you'll hit validation errors immediately. It's the cheap tier; it's supposed to be limited.
• The storyboard math. Segment durations summing wrong is the #1 storyboard failure. The node tells you the exact mismatch, so it's recoverable - just annoying mid-flow.
• Seven references is a lot to feed right. Each frame is a chance to introduce a conflicting detail. Two or three strong, consistent references usually beat seven sloppy ones.
• It's flagship pricing. The badge scales with duration, resolution, model, and audio - this is the most expensive Kling node in the family. Draft the references and prompts first; every render is a line item.

For multi-shot scenes with a consistent lead character, this is the built-in to reach for. Just don't sleepwalk into o1 mode and wonder where your audio went.