Image to Video: a Short Tutorial

By Arcturus

1/12/2026
I think the best way to create a video is to start with a good quality image that has in itself a story to tell. So the video becomes an image animation that can lead to further developments. The single videoclips you can create from a start image are typically 5 or 10 seconds long. For example, I started from the following image: As you can see, there is some kind of spaceship, or flying saucer, just above a swamp filled with strange creatures, under a cloudy, stormy sky on an inhospitable alien planet. To animate this scene I used Kling 2.5 Turbo model with the following prompt: "The spaceship turns on its engines and takes off towards the stormy sky, while the strange alien animals start crawling all around, and the lightnings fall here and there." To make the scene more vibrant I asked the model a soundtrack: "Dramatic space opera orchestra music with angelic choir." And here is the 10 second clip I got at my first attempt: Kling models are quite effective for creating consistent video clips. In case you want to proceed with creating a video longer than 10 seconds, the last frame of the clip can be extracted and used as the starting image for a second clip, with a different prompt, then the two clips can be merged together. However, if you use this process more than once, the quality of the start image deteriorates, and consequently the video quality as well. A better method is to use the last frame as a reference image for a new image created with an image-to-image model, with a prompt containing all the elements needed to continue with the story. Once added to the previous one, the new clip created with this initial image will appear like a scene change in a movie, maintaining the necessary coherence of the story elements.

Tags: ai video, image to video, start image, image to image, video clips