Text or image input
Start from a written shot or a supplied frame.
Kling O3 (Kling 3.0 Omni) on Cinel turns text, first and end frames or ordered references into 3–15 s video at 720p, 1080p or 4K. From 93 credits.
See how a reference frame becomes a moving shot.
Start from a written shot or a supplied frame.
Use first/end frames or an ordered multi-reference set.
Choose 720p, 1080p or 4K for the delivery target.
An engineering and creative breakdown of model performance, architecture, and practical production workflows.
Kling O3 — known in Kuaishou's naming as Kling 3.0 Omni — is the most versatile model in the Kling family. It starts from whatever you have: a written shot, a first frame, a first and an end frame, or an ordered set of reference images. It then delivers clips from 3 to 15 seconds and lets you choose 720p for drafts, 1080p for delivery or 4K when the footage has to hold up on a big screen. On Cinel the default is a 5-second 720p clip in 16:9, with pricing from 93 credits.
O3 combines the widest input range in the Kling lineup with the highest resolution ceiling. Kling O1 is limited to image-to-video at fixed lengths; Kling 3.0 Turbo skips references and 4K. O3 accepts text alone, frame pairs, or several ordered references and still lets you pick any duration within its 3–15 second range.
Technically it sits on Kling's third-generation Omni foundation, which added text-to-video to the Omni line and introduced reference-guided generation. The latest Omni update also improved its editing and reference workflows — including longer 15-second video input and output and 4K source footage — which is why production teams treat O3 as the "escalation" model: when a brief needs consistent characters from references, a precise start and end, or 4K delivery, this is where they go. Sound can be generated with the clip, with a smaller surcharge than on Kling 3.0 at 720p (about 1.33× versus 1.5×).
| Item | Kling O3 |
|---|---|
| Developer | Kuaishou (Kling AI) |
| Also called | Kling 3.0 Omni / V3 Omni |
| Modes on Cinel | Text-to-video, first/end frame, multi-reference |
| Model-level extras | Reference-to-video, video editing (up to 15 s, 4K source) |
| Duration | 3–15 s (Cinel default 5 s) |
| Resolution | 720p / 1080p / 4K |
| Aspect ratios | 16:9, 9:16, 1:1 |
| Price factors (model) | 1080p ≈ 1.33×, 4K = 5× the 720p rate; sound ≈ 1.33× at 720p |
| References on Cinel | Up to 2 images |
| Cinel cost | From 93 credits |
Compared with Kling O1 (from 230 credits), O3 adds text-to-video, flexible 3–15 s durations and a 4K tier, while O1 stays focused on fixed 5/10 s keyframe animation. Compared with Kling 3.0, O3 adds reference-led control and a lower sound surcharge; choose Kling 3.0 when a simple text or image prompt with energetic motion is all you need.
Yes. Kling O3 is the short name for Kling 3.0 Omni, built on Kuaishou's V3 Omni architecture.
A text prompt, a first frame, a first and end frame, or an ordered set of reference images, up to two images in the current panel.
Yes. You can choose 720p, 1080p or 4K. At model level 4K costs five times the 720p per-second rate, so use it for finals.
O3 adds text-to-video, any duration from 3 to 15 seconds and higher resolution options. O1 focuses on image-to-video with first and last frames at 5 or 10 seconds.
The model supports instruction-based editing of existing footage in advanced workflows. On Cinel, check the generator for the modes currently enabled.
It starts from 93 credits. Duration, resolution, sound and references change the total shown before you generate.