1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
const axios = require('axios');
const api_key = "YOUR API-KEY";
const url = "https://api.segmind.com/v1/kling-o3-text2video";
const data = {
"prompt": "A majestic golden eagle soaring over snow-capped mountain peaks at sunrise, cinematic wide angle shot, breathtaking natural scenery, ultra detailed",
"duration": 5,
"mode": "pro",
"generate_audio": false,
"aspect_ratio": "16:9",
"negative_prompt": "blur, distort, and low quality",
"cfg_scale": 0.5,
"shot_type": "customize"
};
(async function() {
try {
const response = await axios.post(url, data, { headers: { 'x-api-key': api_key } });
console.log(response.data);
} catch (error) {
console.error('Error:', error.response.data);
}
})();Attributes
Describe scene as a director: subject, action, camera, and mood.
Output video length in seconds, 3-15.
Allowed values:
Quality tier. Std for fast drafts, pro for production output.
Allowed values:
Define up to 6 scenes for storyboard-style multi-shot videos.
Enables native synchronized audio.
Voice IDs from Kling Create Voice for synchronized speech.
Output video aspect ratio.
Allowed values:
Elements to suppress in the output.
Prompt adherence strength (0–1).
min : 0,
max : 1
Camera shot type for the generation.
To keep track of your credit usage, you can inspect the response headers of each API call. The x-remaining-credits property will indicate the number of remaining credits in your account. Ensure you monitor this value to avoid any disruptions in your API usage.
Kling O3 Text-to-Video — AI Video Generation Model
What is Kling O3 Text-to-Video?
Kling O3 (Video 3.0 Omni) is the flagship text-to-video model from Kuaishou Technology, launched in February 2026. It transforms natural language descriptions into high-quality video clips up to 15 seconds long, with support for native synchronized audio generation in a single inference pass. Built on the "Omni One" architecture, Kling O3 combines 3D Spacetime Joint Attention with Chain-of-Thought reasoning — enabling the model to think through scene composition, motion physics, and camera behavior before rendering a single frame. The result is cinematic video that respects real-world physics and maintains consistent characters across shots.
Unlike earlier text-to-video systems that treated generation as a simple mapping task, Kling O3 operates more like a film director: it interprets your prompt for narrative intent, decomposes it into scene elements, plans motion paths and lighting, and executes the full sequence — all in one generation.
Key Features
- Native audio generation: Dialogue, ambient sound, and music are synthesized simultaneously with the video — no post-processing required. Lip-sync, spatial audio, and breath timing emerge naturally from the same generation pass.
- Multi-shot storyboarding: Generate up to 6 distinct camera shots with individual prompts and durations in a single API call, enabling full scene narratives without manual stitching.
- Physics-accurate motion: Built-in physics engine models gravity, balance, deformation, collision, and inertia for realistic character and object movement.
- Flexible aspect ratios: 16:9 for widescreen, 9:16 for mobile/social, and 1:1 for square content.
- Standard and Pro quality modes: Standard delivers faster turnaround for drafts and iteration; Pro produces cinematic-grade output for final delivery.
- Extended duration: 3 to 15 second clip lengths — significantly longer than most competing models.
Best Use Cases
Content creation and marketing: Generate product showcase videos, brand storytelling clips, and social media content at scale without a full production team. The character consistency features are especially useful for series content requiring the same subject across multiple videos.
Film pre-visualization: Directors and storyboard artists can quickly prototype shot sequences, camera movements, and scene compositions before committing to production shoots.
E-commerce: Demonstrate products in context — lifestyle scenes, environments, and dynamic product shots generated from a brief description.
Education and training: Create explainer videos with synchronized narration, visualize abstract concepts, or produce multilingual training scenarios using the built-in multi-language audio support (English, Chinese, Japanese, Korean, Spanish).
Developers and product teams: Integrate Kling O3 into content pipelines via the Segmind API endpoint at https://api.segmind.com/v1/kling-o3-text2video. Average generation time is approximately 75 seconds per clip.
Prompt Tips and Output Quality
Write prompts as scene directions, not image descriptions. A strong Kling O3 prompt specifies the subject, action, camera movement, lighting, and mood — for example: "Medium shot of a woman walking through a neon-lit Tokyo street at night, camera panning slowly, warm glow reflecting on wet pavement, cinematic and moody."
Use the cfg_scale parameter to control how closely the output follows your prompt: values around 0.5 are a good default; push toward 1.0 for strict adherence. Use negative_prompt to suppress common artifacts — including blur, distort, low quality, shaky camera covers most cases. For Pro mode, generation produces sharper motion and better lighting consistency, especially for longer clips above 8 seconds.
FAQs
What is the maximum video length Kling O3 can generate? Kling O3 supports clips from 3 to 15 seconds in a single generation.
Does Kling O3 generate audio automatically?
Audio generation is optional. Enable the generate_audio parameter to get synchronized dialogue, ambient sound, and music alongside the video. You can also pass custom voice IDs via the voice_ids parameter.
What is the difference between Standard and Pro mode? Standard mode is faster and costs less — ideal for drafts and iteration. Pro mode produces higher-quality cinematic output with better lighting, motion, and detail, particularly noticeable in clips over 5 seconds.
Can I generate multi-shot videos with different scenes?
Yes. Use the multi_prompt parameter to pass an array of scene segments, each with its own prompt and duration. This supports up to 6 camera cuts in a single generation.
What aspect ratios are supported? 16:9 (landscape), 9:16 (portrait/mobile), and 1:1 (square). Select based on your target platform.
How long does generation take? Average latency is approximately 75 seconds. Longer durations and Pro mode will take slightly more time.
Popular Models
