Vidifai

Vidifai turns written or spoken material into a finished video. You give it one of three things — a rough idea, a full script or set of lyrics, or an audio file — and it does the rest: expanding the idea into a narrative, transcribing audio down to word-level timestamps, choosing a visual theme, writing an image prompt for each fragment of text, generating the images, adding narration where you haven't supplied your own, and compiling it all into an MP4 with crossfades and slow pans. In between there is an editor with a live preview, where you can rewrite a line, regenerate an image you don't like, or flip the whole thing to vertical.

It exists because most tools of this kind generate an image per sentence in isolation, so the word "pause" becomes a picture of a pause button and a character's face changes between shots. Vidifai runs a story elaboration pass first, so an abstract line is read in the context of the whole piece, and it writes a description of each recurring character up front so they stay the same person throughout.

It is live, and rendering happens on AWS Lambda so a long video doesn't tie up a server. The paid tier isn't enforced yet — the plan keys are in the code, the last mile of billing isn't done.