Everything HeyGen Video 1.0 can do

HeyGen Video 1.0 is HeyGen's video model. It makes the sound in the same pass as the picture: lines spoken word for word with the mouth in time, the room around them, effects and music. Clips run 5 to 15 seconds at 480p or 768p, in six shapes from 21:9 to 9:16. It can open on a still of yours, or build a new scene around up to nine pictures, three videos and three sound files, twelve in all.

Every clip below was made by the step under it. Change a placeholder and press Gen to make your own.

Quote the line

Put each spoken line in quotes, and say who says it and how. The words come out as written, the mouth in time with them. Speech runs at about two and a half words a second, so count the words before you set Duration: five seconds holds about twelve. For another language, write the line in that language and name it, and leave the rest of the prompt as it is: the baker at the top says the same line in English and in French.

Name the process

For a look other than live action, say how the picture was made: the material, the marks it leaves, a short list of colors, and what it must not be ("no 3D, no gradients"). A style named as a mood comes out as a smooth digital copy. The keeper, the storm and the sound are the same words in all three; only the style line changes. To hold one look across a series, reuse the style line word for word.

Two people talking

For two people, describe each so they can't be mixed up, give each their own line, and say that each mouth moves only on its own line.

Start from a still

Give it a first frame and the clip opens exactly on it, in the still's shape. The still already settles the look, so the prompt only says what moves, what the camera does and what it sounds like. For a talking clip, start from a portrait with the mouth closed. These stills were made with FLUX 3 Image; drop in your own photo instead.

With a product, keep the product still and move the light or something around it: its shape and surface stay as they were photographed.

Say what it sounds like

Every clip comes with sound, and the prompt decides what's on it. End with an Audio: line that names each sound: what's close to the camera, what's across the room, what's far away. Tie an effect to the moment that makes it, the brakes screeching as the train pulls in, and it lands on that moment. Ask for each sound loudly, since a quiet one can get lost, and write "no music" unless you want a score.

Write the whole brief

The model does what the prompt says and fills everything else with its own defaults, so a long, specific brief beats a single line. Open with the camera: the lens, the aperture, what holds it, how much of the frame the subject takes. Words like "cinematic" change little. The same fisherman, on an 85mm and on a 24mm:

Rule out the defaults you don't want, too: faces drift toward a retouched look and blank surfaces pick up made-up logos, so every prompt here says no to both. Prompt rewriting is off, so your brief reaches the model as written; turn it on when all you have is a one-line idea.

Short lettering comes out right when it's in capitals, in quotes, and named as the only lettering in frame. Keep it to a few words: the longer the line, the more letters can slip.

Bring your own pictures

Instead of a first frame, give it pictures to build a new scene around: a person, a product, a place, up to nine. Name each one by what it's for, then say what must stay as it is in its picture. Here the beekeeper and the jar from above meet at a market neither photo showed. A clip built from pictures costs twice as much a second as one from words or a first frame.

To build on any step, press Mix to copy its recipe into your own page.