КАДР

Image-to-video or text-to-video: when to make a video from a photo and when from text

Alen founder of KADR Updated 2026-10-26 3 min read

In text-to-video the model invents the whole video from a description; image-to-video starts from your picture, and the model adds motion. Text gives more freedom and surprises, a photo gives more control: you know in advance what the frame, the hero and the product look like. For ads, products, characters and anything where accuracy matters, the «image first, then video» approach pays off: an image costs about 25 rubles and is easy to redo, while a failed clip costs more. For atmospheric scenes and looking for ideas, video from text works. Below are a comparison and rules for choosing.

The modes in brief

Text-to-video (video from text). You write a description, the model invents everything: the hero’s appearance, the setting, the angle, the light.

Image-to-video (video from a photo). You upload an image, it becomes the first frame. The model adds motion, while you have set the look.

Comparison

From text From a photo
Control over the look of the frame low high
Freedom and unexpected ideas high lower
Product accuracy no, the model invents it yes, your real product
Same character across scenes hard easy, with one image
Attempts until the result more fewer
What you need only text an image plus text about the motion

When to choose what

From a photo:

From text:

  • atmospheric scenes: nature, city, weather;
  • looking for ideas and drafts;
  • abstract and fantastic scenes where the exact look does not matter.

The «image first, then video» approach

For most tasks it pays to split the work:

  1. Make the frame as an image in Nano Banana Pro: hero, product, setting, light. About 25 rubles, quick to redo.
  2. Once you like the frame, upload it to a video model and describe only the motion and the sound.

This way you fine-tune the look on a cheap image rather than on expensive clips.

How to write a prompt for video from a photo

In the prompt, do not describe again what is already in the image. Write only:

  • what moves and how;
  • the camera move;
  • the sound;
  • what must stay unchanged.
The woman slowly turns her head toward the camera and smiles, a light breeze moves her hair. Static camera. The face stays the same. Ambient: quiet street sounds.

The less motion, the less the model changes the source. Ready prompts: Kling 3.0, Veo 3.1.

How to write a prompt for video from text

Describe everything: shot and camera, hero, action, place, light, style, sound. In detail: how to write video prompts.

What it costs

Step Model Approximately
First frame Nano Banana Pro, 2K ~25 rub.
Video 6 s, draft Seedance 1.5 Pro, 480p with sound ~30 rub.
Video 6 s, 1080p with sound Veo 3.1 Fast ~126 rub.
Video 6 s, people Kling v3, 1080p with sound ~168 rub.

Prices on KADR (kadr-ai.ru), October 2026. This is not a public offer.

In short

Video from text gives freedom, video from a photo gives control. For ads, products, characters and animating photos choose image-to-video: first an image for 25 rubles, then the clip. For atmospheric scenes and looking for ideas, text-to-video will do. In a prompt for video from a photo, describe only the motion, the camera and the sound.

Frequently asked questions

How is image-to-video different from text-to-video?
In text-to-video the model creates the whole frame from a text description. In image-to-video your picture becomes the first frame, and the model adds motion. The second mode gives more control over what the result looks like.
Which is better for a product ad?
Video from a photo. The product has to be exactly yours, so take its real photo, change the background in the image if needed, and then animate it.
Which is better for a character in several scenes?
Video from a photo or with a reference. First make an image of the hero, then use it as the first frame or a reference in every scene. That way the hero stays the same.
When should I choose video from text?
For atmospheric scenes (nature, city, abstraction), looking for ideas, and cases where the exact look of the frame does not matter.
Why is video from a photo cheaper in practice?
An image costs about 25 rubles and is quick to redo, while a clip costs from several dozen to hundreds of rubles. Once you have worked out the frame on an image, you spend fewer attempts on the video.

Images and video in one window

On KADR: Nano Banana Pro for the first frame and Veo, Kling, Seedance for video. The ready-made «Animate photo» idea on the start screen walks you from image to clip step by step. Pay in rubles, tokens never expire.

Try it
Alen
Founder of KADR. I can see from users' generations which mode leads to fewer failed attempts.