What you need
- A browser, a computer or a phone.
- An idea for one scene: 5 to 10 seconds.
- 10 minutes.
No camera, editing software or skills required.
Step 1. Choose: from text or from a photo
From text (text-to-video). You describe the scene in words, and the model invents everything itself. More freedom, more surprises.
From a photo (image-to-video). You upload a picture, it becomes the first frame, and the model adds motion. The result is more predictable. For a beginner it is easier to start here. In detail: how to animate a photo with AI.
Step 2. Choose a model
For your first clips, take an inexpensive model: you are learning to write prompts, and failed attempts are inevitable.
| Model | What for | 6 seconds on KADR |
|---|---|---|
| Seedance 1.5 Pro, 480p | the cheapest tests | ~30 rub. |
| Veo 3.1 Lite, 720p | tests with sound | ~52 rub. |
| Veo 3.1 Fast, 1080p | a good result at a low price | ~126 rub. |
| Kling v3, 1080p | people and camera movement | ~168 rub. |
KADR prices with sound, October 2026. This is not a public offer. Once you get the hang of it, see the comparison of all models: which AI to choose for video.
Step 3. Write a prompt
The formula for your first clip:
Who or what is in the frame + one action + where + how the camera moves + light + sound
Three ready prompts to start with:
-
Nature:
A ginger cat sleeps on a windowsill, snow falls slowly outside the window, slow camera push-in, warm light from the room. Sound: quiet purring.
-
City:
An evening street after the rain, neon signs reflected in the puddles, a person with an umbrella walks past, static camera. Sound: rain and city noise.
-
Food:
Close-up of a cup of coffee on a wooden table, milk is poured into the coffee, a pattern spreads across the surface, slow camera movement from the side. Sound: quiet café music.
Rules:
- one action per scene;
- one camera movement;
- specific words instead of «beautiful» and «cinematic».
If you are not sure about the wording, press the «Improve» button on KADR: the AI will rewrite the prompt and suggest a model and settings. It is free up to 10 times a day, and up to 3 times with a zero balance. More on prompts: how to write prompts for video.
Step 4. Settings
- Duration: 5 or 6 seconds to start with.
- Format: 9:16 for Reels and Stories, 16:9 for YouTube and screens, 1:1 for the feed.
- Resolution: the minimum for tests, 1080p for the final version.
- Sound: turn it on if you need speech or atmosphere. If music will go on top, it is cheaper without sound: almost half the price on Seedance 1.5 Pro.
Step 5. Run it and evaluate
Generation takes from a minute to a few minutes. Look at the result and decide:
- All good: make the final version in 1080p.
- Wrong motion: simplify the prompt, keep one action.
- Wrong look of the frame: start from a photo instead of text.
- The model refused: switch the model or remove real people and brands from the prompt. More: the AI refuses to generate.
Typical beginner mistakes
- Going straight for a flagship in 1080p. Expensive attempts get burned on learning. Learn on a cheap model.
- A whole plot in one prompt. «A girl walks into a café, sits down, orders and laughs» will be ruined by the model. That is 3 scenes.
- Expecting a film from one generation. One prompt is one short shot. A long video is assembled from shots.
- Cropping a horizontal video into a vertical one. Set the format before generating.
What next
- Put together a Reel from several scenes: how to make Reels with AI.
- Make a cartoon: how to make an animated film with AI.
- A product ad: an AI product ad video.
In short
A video with AI takes 10 minutes: choose «from a photo» or «from text», an inexpensive model, write a prompt with one action and one camera movement, set the duration and format. Make your first clips on Seedance 1.5 Pro or Veo 3.1 Lite in low resolution, and the final version on the model that suits the task in 1080p.
