Video · Kuaishou
Kling O3 (Omni)
The omni version of Kling 3.0, built to keep a character or product identical across clips. Pick it when the same face must come back shot after shot.
Anonymized Low costWhat sets it apart
Kling O3 (3.0 Omni) is built around references: you feed it images, elements, video and a voice and it keeps the same character and voice across scenes. Pick it over plain V3 when consistency across shots matters more than camera control.
- Reference-to-video with elements
- Voice binding keeps a speaker consistent
- 3 to 15 s, multi-shot storyboard
Recommended for: keep the same character or product across shots.
Strengths and limits
Where it shines
- References as elements: up to 4 subjects, each with extra angles
- Up to 4 scene images to fix the set, light and style
- Sound and dialogue generated with the picture
Think twice when
- Name each reference in the prompt by its tag (@Element1, @Image1)
- 4K is among the dearest options per second
- 3 to 15 s per clip, native sound
- Text, image or references to video
Use cases
Recurring mascot
Using the attached mascot images as references, show the same fox mascot waving in a bakery, then walking down a rainy street. Keep its red scarf and round glasses identical.
Reference mode locks the look of a character across shots.
Two-person dialogue
Two friends at a cafe table. The woman says 'Did you book the tickets?' and the man answers 'Yes, we leave at nine.' Natural lip-sync, quiet cafe ambience.
Per-character lip-sync and voice binding fit conversations.
Start and end frame
Animate from my first frame (empty desk) to my last frame (desk covered in notes) over 8 seconds, with a time-lapse feel and a pencil scratching sound.
It supports first and last frame control, plus references.
Against its rivals
- Kling V3V3 favors camera and motion control, O3 favors identity and references.#12 of 22 on Artificial Analysis Image to Video Arena (1,055 Elo)
- Seedance 2.5Seedance 2.5 takes up to 30 images, 10 videos and 10 audio clips.#2 of 18 on Artificial Analysis Text to Video Arena with audio (1,142 Elo)
- Gemini Omni FlashOmni Flash edits video through conversation and extends scenes.#7 of 18 on Artificial Analysis Text to Video Arena with audio (1,113 Elo)
Version history
Each version, dated, with what it changed over the one before.
- Kling 2.6Earlier, not in Sub Rosa
- First Kling with native audio, 5 or 10 s clips.
- Kling 3.0 Omni (O3)Latest
- Unified prompting where text, images, video and audio are equal inputs.
- Voice binding and video-based character references keep identity across scenes.
- Multi-shot storyboard up to 15 s with per-character lip-sync.
Benchmarks
How to read these scores: what a benchmark tells you.
Artificial Analysis Image to Video Arena: #14 of 22 models measured in the catalog. The same blind duel, starting from one image: which clip brings it to life better. An Elo rating: higher is preferred more often. 100 points apart means the higher-rated model wins about 64% of duels.
- 1. MiniMax H3 MaxMiniMax H3 (Hailuo)1,195
- 2. MiniMax H3MiniMax H3 (Hailuo)1,181
- 3. Wan 3.0Wan 3.01,164
- 4. HappyHorse 1.1HappyHorse1,104
- 5. Grok Imagine 1.5Grok Imagine video1,098
- 6. HappyHorse 1.0HappyHorse1,083
- 7. Veo 3.1 Full QualityVeo 3 and 3.11,082
- 8. Wan 2.7Wan 2.x1,077
- 9. Grok ImagineGrok Imagine video1,072
- 10. Veo 3.1 FastVeo 3 and 3.11,066
- 14. Kling O3 ProKling O3 (Omni)1,044
- 16. Kling O3 StandardKling O3 (Omni)1,036
See the numbers
| Model | Elo |
|---|---|
| 1. MiniMax H3 Max | 1,195 |
| 2. MiniMax H3 | 1,181 |
| 3. Wan 3.0 | 1,164 |
| 4. HappyHorse 1.1 | 1,104 |
| 5. Grok Imagine 1.5 | 1,098 |
| 6. HappyHorse 1.0 | 1,083 |
| 7. Veo 3.1 Full Quality | 1,082 |
| 8. Wan 2.7 | 1,077 |
| 9. Grok Imagine | 1,072 |
| 10. Veo 3.1 Fast | 1,066 |
| 14. Kling O3 Pro | 1,044 |
| 16. Kling O3 Standard | 1,036 |
Artificial Analysis Text to Video Arena with audio: #12 of 18 models measured in the catalog. The duel for clips with sound: picture, voices, effects and how well they match. An Elo rating: higher is preferred more often. Only models that make their own sound take part.
- 1. Wan 3.0Wan 3.01,156
- 2. Seedance 2.5Seedance 2.51,142
- 3. MiniMax H3MiniMax H3 (Hailuo)1,138
- 4. MiniMax H3 MaxMiniMax H3 (Hailuo)1,134
- 5. Flux 3FLUX 3 video1,127
- 6. Seedance 2.0Seedance 2.0 and 1.51,119
- 7. Gemini Omni Flash 1.1Gemini Omni Flash1,113
- 8. Grok Imagine 1.5Grok Imagine video1,044
- 9. HappyHorse 1.1HappyHorse1,043
- 10. Wan 2.7Wan 2.x1,029
- 12. Kling O3 ProKling O3 (Omni)1,017
See the numbers
| Model | Elo |
|---|---|
| 1. Wan 3.0 | 1,156 |
| 2. Seedance 2.5 | 1,142 |
| 3. MiniMax H3 | 1,138 |
| 4. MiniMax H3 Max | 1,134 |
| 5. Flux 3 | 1,127 |
| 6. Seedance 2.0 | 1,119 |
| 7. Gemini Omni Flash 1.1 | 1,113 |
| 8. Grok Imagine 1.5 | 1,044 |
| 9. HappyHorse 1.1 | 1,043 |
| 10. Wan 2.7 | 1,029 |
| 12. Kling O3 Pro | 1,017 |
All 9 models in this family
| Model | Use | Privacy | Price |
|---|---|---|---|
Kling O3 Standardkling-o3-standard-text-to-video | Text to video | Anonymized | $0.0549 per second of video$0.274 a 5 s clip |
Kling O3 Standardkling-o3-standard-image-to-video | Image to video | Anonymized | $0.0549 per second of video$0.274 a 5 s clip |
Kling O3 Standard R2Vkling-o3-standard-reference-to-video | References to video | Anonymized | $0.0549 per second of video$0.274 a 5 s clip |
Kling O3 Prokling-o3-pro-text-to-video | Text to video | Anonymized | $0.0692 per second of video$0.346 a 5 s clip |
Kling O3 Prokling-o3-pro-image-to-video | Image to video | Anonymized | $0.0692 per second of video$0.346 a 5 s clip |
Kling O3 Pro R2Vkling-o3-pro-reference-to-video | References to video | Anonymized | $0.0693 per second of video$0.347 a 5 s clip |
Kling O3 4Kkling-o3-4k-text-to-video | Text to video | Anonymized | $0.208 per second of video$1.039 a 5 s clip |
Kling O3 4K R2Vkling-o3-4k-reference-to-video | References to video | Anonymized | $0.208 per second of video$1.039 a 5 s clip |
Kling O3 4Kkling-o3-4k-image-to-video | Image to video | Anonymized | $0.208 per second of video$1.039 a 5 s clip |
Specifications
| Specification | Kling O3 (Omni) |
|---|---|
| Released | February 2026 |
| Weights | Closed |
| Accepts | text, images, video, audio |
| Produces | video, audio |
| Clip length | 3 to 15 s, with sound |
| Price | $0.0549 per second of video |
| Privacy | Anonymized |
Prices, context and lengths from the live catalog on October 16, 2026; parameters and licenses from the maker.
Sources
Everything on this page comes from these pages, read on October 16, 2026 or as dated on each score.