Kling 3.0 Omni AI Video Generator

What Kling 3.0 Omni is

• Kling 3.0 Omni is the reference version of Kuaishou's Kling 3.0 series. Kling's own guide says it upgrades Kling VIDEO O1: it reads images, videos and "elements" (a saved subject built from several angles) as part of the prompt and keeps each subject's look while the camera moves, now for up to 15 seconds in one generation instead of O1's 10.
• The guide lists native audio, start and end frames, custom multi-shot with a length and camera for each shot, video elements from a 3- to 8-second clip, and a voice that can be bound to a character. It accepts up to 7 images and runs in 720p and 1080p modes. Kling 4 followed in October 2026; Kling 3.0 Omni stays the cheaper of the two.

Kling 3.0 Omni on RefVid: what you can set

• RefVid runs Kling 3.0 Omni through fal.ai: the Standard endpoint for 720p and the Pro endpoint for 1080p. The photos you add are saved as one subject, so every clip starts from the same face or the same label, and you can bring up to 3 saved subjects into one scene.
• In the box you pick 3 to 15 seconds, 16:9, 9:16 or 1:1, 720p or 1080p, and sound on or off. With no photos it makes a clip from your prompt alone. It takes photos of real people, and an end frame if you want the clip to land on a set picture.

Kling 3.0 Omni prompt guide: one subject, one action per shot

• Name the subject, say where it is, give it one action and one camera move. Kling 3.0 Omni follows a short scene better than a list of five things. The clip above came from three photos of the bottle and a prompt with a place ("a white marble bathroom shelf next to a small green plant"), the light ("morning sunlight falling through a frosted window") and one camera move ("Slow push-in toward the bottle until the LUMEN label fills the centre of the frame, the label text stays crisp.").
• For more than one beat, split it into shots instead of one long sentence. The skincare template runs two shots in one 5-second clip: "Extreme macro: a single glossy golden drop of serum falls from a glass dropper and lands on a smooth pale stone surface", then the product "stands on the pale stone next to a sprig of green leaves in soft morning window light, the camera pushing in gently until the label fills the frame." Each shot prompt can be up to 480 characters, the whole prompt 2,500.

• How to write scene prompts for reference video→

Make a Kling 3.0 Omni video in three steps

• Add 1 to 10 photos of your product or character in the box at the top of this page, or start from a template in the row above it. The box is already set to Kling 3.0 Omni, 720p and no sound.
• Write what happens in a sentence or two. Settings changes the length (3 to 15 seconds), the shape, 1080p and sound.
• Press Generate. If you are not signed in, you sign in first and the clip starts right after. Most of RefVid's template examples were made on Kling 3.0 Omni.

Kling 3.0 Omni pricing on RefVid

• A 5-second clip at 720p without sound costs 155 credits, about half of Kling 4 for the same length; sound and 1080p cost more per second, and the box shows the total before anything runs. New accounts get 180 free credits, which cover a first 5-second Kling 3.0 Omni clip with no card and no watermark.
• After that, one video from $2.99, a one-time credit pack from $9.90 or a monthly plan from $10. Credits never expire, and a failed clip gives its credits back automatically.

• How RefVid pricing works→
• Plans and credit packs→

Kling 3.0 Omni next to every RefVid model: measured clips, wait and cost

• These numbers are our own, not the vendor's: every clip people outside our team started on RefVid from October 27 – November 6, 2026, split by the model they picked, with how many finished and the median wait from the start of the generation to the finished clip. Next to them are each model's photo limit, clip length and what a 5-second clip costs today, so Kling 3.0 Omni can be read against the other models in one table.
• The finished share and the wait are shown only where a model had at least 20 finished or failed clips in the period; the models added most recently have fewer so far. A failed clip costs nothing: its credits go back to the account. The table follows the newest snapshot of our usage statistics.

• AI product video statistics: success rate by model→

When Kling 4 or another model fits better

• Pick Kling 4 for a single take longer than 15 seconds or the closest likeness of a face; it always adds sound and costs about twice as much per second. For a long product clip, Wan 3.0 and Seedance 2.5 also reach 30 seconds. Your saved subject works with every model, so you can run the same photos on Kling 3.0 Omni first and on Kling 4 if the shot needs it.

• Kling 4 vs Kling 3→
• Kling 4 Omni Reference→
• Wan 3.0 AI video generator→
• Reference Image to Video→