Premiumvideo
Veo 3.1
Google DeepMind's state-of-the-art video generation model featuring native audio synthesis, up to 4K resolution, and cinematic realism with advanced physics simulation.
text-to-videoimage-to-videohigh-quality
9クレジットから
KuaishouNew
Kling 2.1
Kuaishou's cinematic AI video model powered by 3D spatiotemporal attention — delivering industry-leading physics simulation, hyper-realistic facial expressions, and up to 1080p output across Standard, Pro, and Master tiers.
11クレジットから
1080p (Pro & Master) / 720p (Standard)
3–8 minutes (peak times may be longer)
Kling 2.1でできること
3D Spatiotemporal Attention
Diffusion-Transformer architecture with 3D spatiotemporal joint attention produces physically accurate motion that respects real-world dynamics
Hyper-Realistic Facial Expressions
Best-in-class character animation with nuanced, life-like facial expressions and precise body movement rendering
Advanced Camera Controls
Cinematic pan, tilt, roll, and zoom controls with spatio-temporal precision for professional-grade cinematography
Three Quality Tiers
Standard (720p) and Pro (1080p) for image-to-video; Master (1080p) for text-to-video with advanced physics
Negative Prompts & CFG Scale
Exclude unwanted artifacts with negative prompts and fine-tune prompt adherence with CFG scale (recommended 0.3–0.7)
Sound Effect Generation
Master tier includes built-in sound effect generation to add synchronized audio to your videos
サンプルギャラリー
About Kling 2.1
Released in May 2025 to mark the first anniversary of Kling AI, Kling 2.1 is Kuaishou's most capable video generation model to date. Built on a Diffusion-Transformer architecture with a 3D Spatiotemporal Joint Attention Mechanism, it models complex motion, object interactions, and scene dynamics with a physical accuracy that outpaces most competing tools. Where many AI video generators produce plausible-looking but physics-defying results, Kling 2.1 is engineered to understand how objects, people, and environments actually move — making it the standout choice for any content that demands realism.
Standard vs Pro vs Master
The three tiers share the same core architecture but differ in input mode, resolution, and feature set:
| Tier | Input Mode | Resolution | Best For |
|---|---|---|---|
| Standard | Image-to-video | 720p | Cost-efficient drafts, social content, rapid iteration |
| Pro | Image-to-video | 1080p | Professional content, marketing videos, enhanced fidelity |
| Master | Text-to-video | 1080p | Cinematic productions from a text prompt, advanced physics |
Standard and Pro are image-to-video models: you supply a reference image and a motion prompt, and the model animates it with natural physics. Master works from text alone, generating high-fidelity cinematic video purely from your description — no input image required. Master is also the only tier with built-in sound effect generation, a significant differentiator since most AI video tools produce silent output that requires separate post-production audio work.
Key Features
- Image-to-video (Standard & Pro) — Upload a reference image as the starting frame; the model animates it while preserving fine details and textures
- Text-to-video (Master) — Describe any scene in text and get a physics-accurate, 1080p cinematic clip without needing a source image
- Motion Brush — Paint which objects should move and control direction and intensity while keeping other elements static
- DeepSeek prompt assist — AI-generated prompt suggestions based on your scene description, theme, or rough idea
- Flexible duration — 5-second or 10-second clips, chainable for longer sequences
- Aspect ratios — 16:9 (landscape), 9:16 (portrait/Reels), and 1:1 (square feed)
- CFG Scale — Adjustable prompt adherence from 0 to 1 (recommended 0.3–0.7)
Tips for Best Results
- For Standard/Pro (image-to-video): Use high-quality source images with clear subjects and good lighting — the model preserves and animates fine details from your reference frame
- For Master (text-to-video): Write motion-specific prompts that describe movement, camera angle, and atmosphere (e.g., "slow dolly push toward subject, sunlight streaking through leaves, cinematic atmosphere")
- Leverage negative prompts — values like
blur, distort, morphing, erratic motion, low quality, artefactssignificantly reduce common AI video artifacts - CFG scale 0.3–0.7 is the practical sweet spot: lower values allow more creative motion interpretation, higher values enforce tighter prompt adherence
- Prototype on Standard, deliver on Pro — for image-to-video work, use Standard to validate motion direction, then re-run on Pro for final 1080p output
- Chain generations for longer videos — use the end of one generation as the start of the next to build coherent sequences beyond 10 seconds
技術仕様
最大解像度1080p (Pro & Master) / 720p (Standard)
最大時間10 seconds per generation (extendable via chaining)
アスペクト比16:9, 9:16, 1:1
生成速度3–8 minutes (peak times may be longer)
出力形式MP4
Model Variants
Kling V2.1 Standard
image to video
Kling V2.1 Pro
image to video
Kling V2.1 Master
text to video
クレジット料金
| Variant | クレジット | Duration |
|---|---|---|
| Kling 2.1 Standard | 11 | 5s |
| Kling 2.1 Pro | 21 | 5s |
| Kling 2.1 Master | 67 | 5s |
1クレジット = $0.012
ユースケース
Character & Narrative Content
Animate human subjects with photorealistic expressions and motion using Standard or Pro image-to-video — ideal for storytelling and avatar creation
Product & E-commerce Video
Bring still product photography to life with natural motion and environmental context for ads, landing pages, and social media
Cinematic Pre-production
Use Master text-to-video to rapidly prototype scenes, camera moves, and visual styles before full production
Social Media Content at Scale
Generate engaging short-form clips — Standard and Pro from a reference image, Master from a text prompt
類似モデル
Premium
Popularvideo
OpenAI
Sora 2
OpenAI's flagship video-and-audio generation model with advanced physics simulation, native synchronized audio, and multi-shot scene control — released September 30, 2025
text-to-videoimage-to-videocinematic
5クレジットから
video
MiniMax
Hailuo
MiniMax's Hailuo 02 video generation models deliver cinematic-grade physics simulation, expressive character motion, and versatile stylization across text-to-video and image-to-video workflows.
text-to-videoimage-to-videofast
13クレジットから