Kling Omni O3 Guide: Image to Video With Multi-Prompt Segments

A florist trims the stems, builds the bouquet, then holds it up to the camera. With a single prompt, you'd cram all three actions into one sentence and hope the timing works. Kling Omni O3 lets you give each action its own prompt and its own share of the seconds, inside one 3–15 second clip that starts from your first frame.

It runs in the Spicy Studio Video Lab (open it with O3 selected) in two tiers, Standard and Premium, and also works as a plain image-to-video model with a single prompt.

First frame: the image the clip starts from. Spicy Studio runs O3 as an image-to-video model, so bring one.

Last frame (optional): the image the clip should finish on.

Duration: any whole number of seconds from 3 to 15; 5 by default.

Shape: taken from the first frame. The aspect ratio menu is switched off for this model.

Generate Sound: off by default; adds synchronized audio at a higher rate.

Multi prompt (optional): a chain of segments, each with its own prompt and length.

Three rules decide what you can build:

Every segment needs a prompt and lasts at least 3 seconds; its length menu runs from 3 to 15.

Segment lengths must add up to the clip's duration. Set the duration slider first, then divide it. If the numbers don't match, the Video Lab shows the target and the current total, and won't generate until they agree.

The clip tops out at 15 seconds. You can add up to 8 segments, but only 5 fit: five 3-second segments fill all 15 seconds.

So two segments need at least a 6-second clip, three need 9, four need 12 and five need the full 15.

First frame: a florist in her 30s at a wooden workbench, eucalyptus and peonies laid out, soft window light, cropped to 9:16. Set the duration to 12 seconds, then add:

Segment 1, 4 seconds: “Close on the florist's hands trimming eucalyptus stems with shears. Static camera, soft window light.”

Segment 2, 5 seconds: “The florist gathers peonies and eucalyptus into a loose bouquet and wraps it in brown paper. The camera eases back to a medium shot.”

Segment 3, 3 seconds: “The florist ties the twine, lifts the bouquet toward the camera and smiles.”

4 + 5 + 3 = 12, so the Video Lab accepts it. It costs 168 credits on Standard (12 × 14), 228 on Premium or on Standard with sound, and 288 on Premium with sound.

Each segment is prompted on its own, so refer to people and objects the same way every time, give each segment one action and say what the camera does. Three seconds suits a single gesture; give the most time to the beat that carries the clip.

The first frame fixes the look and, since O3 takes its shape from it, the format. Crop it before uploading: 9:16 for Reels, TikTok and Shorts, 16:9 for YouTube, 1:1 for a square post.

The last frame tells the model where to land: a product clip that must end on the packshot, a character who should finish in a set pose, a shelf that goes from empty to stocked. With segments, the last frame is where the final segment ends, so write that segment toward it.

No frames yet? Make them in the Image Lab. Nano Banana 2.1 creates a 2K still for 12 credits; give it the first frame as a reference so both frames share the same person and setting.

O3 bills per second of clip, and segments don't change the price.

Standard: 14 credits a second, 19 with sound. 5 seconds: 70, or 95 with sound. 15 seconds: 210, or 285.

Premium: 19 credits a second, 24 with sound. 5 seconds: 95, or 120 with sound. 15 seconds: 285, or 360.

Draft on Standard with sound off, then switch to Premium or turn sound on for the version you'll publish. Each run is a new generation, so the final won't copy the draft exactly; the draft tests your frames and prompts.

The shortest clip, 3 seconds on Standard, costs 42 credits, more than the 25 free credits on a new account. The €10 Spicy Basic plan's 600 monthly credits cover eight 5-second Standard drafts; see pricing.

A single 5- or 10-second shot: Kling 3.0 costs less per second (12 on Standard), offers 16:9, 9:16 and 1:1, and renders 1080p on Pro (Kling 3.0 guide).

Past 15 seconds, or guided by reference images, video or audio: Seedance 2.5 runs 4–30 seconds with audio included in the price (Seedance 2.5 guide).

Text only: Seedance 2.5, Seedance 2.0 or Veo 3.1.

Moves copied from real footage: Kling 3.0 Motion Control in Character Swap (motion transfer guide).

For specs and credits on all three, read Seedance 2.0 vs Kling O3 vs Kling 3.0.