It starts with a script
You type what the video is about and pick a genre. The genre fixes the running time, so there is no length to choose. Claude first finds an idea in your sentence, then writes the video scene by scene: what is said, what is on screen and how the camera moves.
If you would rather have your own story told, the plot setting keeps to your brief event for event, with the characters you named and the lines you quoted.
Then a voice and a picture for every scene
ElevenLabs reads the script with a timestamp on every word, in a voice chosen for the genre and the language unless you pick another. Long pauses are cut out of the reading, and each scene is timed to its own phrase.
GPT Image draws a frame for each scene. A video model then animates each frame into a clip exactly as long as the shot that needs it.
Last comes the cut
FFmpeg joins the clips, evens out their light and colour so that the cuts do not jump, lays music under the voice, burns in the captions word by word and masters the sound. The file that comes out carries “AI-generated” in its own metadata.
All of this takes minutes rather than seconds. The editor shows you each stage while it runs, and you can leave the page and come back.