Advertising
Product ad
A car races through a night city. Neon signs, wet asphalt, a sense of speed.
- Tools /
- AI video generator
AI video generator
Video generation with artificial intelligence (Kling v3, Kling v3 Pro and VEO 3.1 models)
See what you can create
Videos made by the generator from a text prompt.
Nature
Wildlife shot
A cheetah runs across the savannah. The camera tracks alongside, natural light, realistic footage.
Cinematic
Cinematic chase
Two red supercars on a wet track at dusk. The camera low at the side, headlight flares.
Your video
Not started yetYour video will appear here
Pick an example above or describe your own idea. Once started, the result shows up in this panel.
For a short clip pick one main action. Split a complex story into several scenes.
Generation history
Your results in one placeHistory is saved for registered users only
You can generate a video without an account, but the clip is gone once you close the page. Sign in to keep your results here.
How to use the generator
How to create a video: from description to result
Here are tips for your first run and techniques for those who already work with neural networks. Start with the section you need.
How to choose a model
The model is the neural network that will create the clip. It defines the available settings, duration and price.
Kling v3A good place to start
One subject, simple camera motion, product shots or animating a photo.
- 3–15 seconds
- 16:9, 9:16, 1:1
Kling v3 ProTo compare results
Use the same prompt if the Kling v3 result lacked detail or the motion came out imprecise.
- 3–15 seconds
- 16:9, 9:16, 1:1
VEO 3.1 FastWhen key frames matter
Lets you set the start and end of the clip with images, or pass the look through style references.
- 8 seconds
- 16:9, 9:16
How to write the video description — the prompt
The prompt is the text the neural network builds the video from. Say what is in the frame and what happens. Add the place, camera motion and light if they matter to your idea.
1. SubjectWho or what→2. ActionWhat they do→3. PlaceWhere and when→4. CameraAngle and motion→5. Light and styleHow the shot looks
A cheetah runs across the African savannah. The camera moves alongside at the animal level. Soft morning light, realistic footage.
SubjectA cheetahActionRunsPlaceAfrican savannahCameraMoves alongsideLight and lookSoft light, realistic
Useful camera words
- Dolly-in
- The camera moves closer to the subject and the perspective changes.
- Pan
- The camera stays in place and turns to the side.
- Tracking shot
- The camera follows a person, an animal or a car.
- Static shot
- The camera stays put, only the objects in the frame move.
Check before you start
- The clip has one main action
- It is clear where and how fast everything moves
- The time of day or the light source is named
- The description has no contradicting instructions
If the idea does not fit one action, split it into several clips.
How to animate an image
The photo defines the look of the scene. In the description say what should move and what must stay.
1
Pick a sharp photo
The main subject must be clearly visible and important details must not be cropped.
2
Describe the motion
Start with one or two actions: "turns their head", "the camera moves closer".
3
Say what to keep
For example: "keep the facial features, clothing and background of the source image".
For VEO 3.1 FastThe first and last frames set how the clip begins and ends. Style references help pass the look of the scene. In VEO these are separate modes: use either the frame pair or up to three style references.
Settings in plain words
For the first run the model, duration and aspect ratio are enough. The other parameters help fine-tune the result.
- Duration
- With Kling start at 5 seconds. VEO creates 8-second clips. A long scene is better split into several short ones.
- Aspect ratio
- 16:9 for a site and YouTube, 9:16 for Shorts and Reels, 1:1 for feed posts.
- Prompt adherence · Kling
- A CFG value from 0 to 1 sets how strictly the model follows the description. Start at 0.5 and change it gradually: a higher value does not always improve the video.
- Negative prompt · Kling
- A separate field for what must not appear in the frame. For example: "blur, captions, jerky motion, distorted objects". Start with a few specific limits.
- Seed
- Seed is a number that sets the initial randomness of the generation. Save it to compare variants while the other settings stay the same. On its own it does not lock a character look.
Frequently asked questions
What to try if the first result is not what you expected.
Can I write the prompt in Russian?
Yes. If the model misses details, turn on the translation into English. Write a clear prompt first — a translation will not fix a contradictory description.
How long should the description be?
Start with 2–4 sentences: who is in the frame, what they do, how the camera moves and what the light is. Add only the details that must be visible in the clip.
What if the motion looks unnatural?
Keep one main action, slow it down and remove extra objects. When animating a photo, take a cleaner shot where the subject is not covered by other objects.
How do I keep the look of a subject or character?
Upload the first frame and describe mostly the motion. Use the same source for related scenes. Exact preservation of small details is not guaranteed — check the result before you use it.
Are limits charged if the video fails?
No. Limits for a failed generation are refunded. If the safety filter rejected the request, change the description or the uploaded images and run the task again.
🍪 By using this website, you agree to the processing of cookies and collection of technical data to improve website performance in accordance with our privacy policy.