Video generation with a neural network: the complete guide

A neural network draws a picture in seconds, and we're used to that by now. Video is the next frontier: the same models now don't just draw a frame but make it move. An old photo blinks and smiles, one sentence of text turns into a five-second clip, and a character who never existed speaks in your voice. This is video generation — and in 2026 it has gone from "wow demo" to a working tool.

This guide is a connected conversation, not a list of services. From the key question "how does moving video come out of text or a photo at all" to concrete tasks: animate an old photo, shoot a clip from a description, make a talking avatar-narrator, translate and voice someone else's video or remove an unwanted object from your own.

Describe a short scene or upload a photo right here, in the Twelver chat — and watch it come to life. Video is noticeably more expensive than pictures, so the first clips aren't handed out "to everyone at once": sign up and complete a couple of onboarding steps — they grant starter tokens, enough for your first generations.

Why video is a separate story (and why it costs more)

Honestly from the start: generating video is dozens of times "heavier" than a picture. The network draws not one frame but tens per second, and keeps the face, light and physics of motion consistent between them. So a clip both takes longer to compute and costs more — that's not marketing but the arithmetic of computation.

The practical takeaway for you: don't waste generations. This guide is built so that you get the result you need on the first or second try, rather than burning tokens on a lottery. Wherever it matters, we explain how to compose the shot in advance so you don't have to reshoot.

What's already real, and what isn't yet

Real today: short clips (usually 5–10 seconds) of high quality, animating photos, talking avatars, translation and voicing. People build ads, social content, avatar-hosts and bring family archives to life on top of this.

Still limited: long scenes with a plot, perfect face stability over minutes, complex physics (hands, a crowd, text in the frame). The technology moves fast — what's "almost" today becomes the norm in six months. So this guide is living: we update the breakdowns as new models come out.

Contents

1. 1.Animate a photo: how to make a video from a photograph
2. 2.Text to video: a clip from a single description
3. 3.A talking avatar: how to make a face speak
4. 4.The best neural network for video
5. 5.Sora: what it is and how to get access
6. 6.Kling: how to use a neural network for video
7. 7.Runway and Pika: control and speed
8. 8.Automatic subtitles for video
9. 9.Translating and dubbing video with a neural network
10. 10.Remove an object from video with a neural network
11. 11.Video for Reels, Shorts and TikTok with a neural network
12. 12.An ad video from a product photo
13. 13.Video for listings on Amazon and other marketplaces
14. 14.Real-estate video from photographs
15. 15.50 video prompts you can copy