AI NewsDev toolsAnnouncement

AI Video Factory turns one topic into a Claude-made YouTube video

AI Video Factory is a Claude Code project. It takes one topic and returns a narrated YouTube video. The repository has 518 stars in 20 days.

AI News

Editorial2 min read

LinkedInX
GitHub social preview for ersinkoc/AI-Video-Factory

Image: GitHub

Why it mattersA marketer asks for a video and gets a plan, a script and a finished file in the same tool. The human picks the style and reviews. The agent does the rest.

A marketer gives a topic and gets a finished YouTube video. The script, the shots, the art and the captions are all picked by Claude agents. A Studio in the browser keeps a human in charge of the brief and the review. The project is called AI Video Factory. It is open source under an MIT licence.

Ersin Koç released the project on GitHub on 29 October 2026. The repository has 518 stars and 71 forks. It holds 244 commits since release. The README says the Claude agents do four jobs. They write the story and script, plan the shots, design the vector art, and review the result. The Studio runs in the browser at a local address. It has no login.

What the agents do

The project splits the work into four jobs. One agent writes a story and the voice-over script. A second plans the shots, scene by scene. A third designs the vector art and the characters. A fourth reviews the result. The author's goal is that a single topic becomes a published video without a person running each stage.

A video can be built in four visual modes. The first is rigged vector animation. The second uses still images with motion. The third mixes stills and AI-generated clips. The fourth is fully generated footage. The CLI takes a --minutes flag, so the length is up to the person who runs it.

What is in the box

The README lists 64 style packs. Each pack is a palette, drawing rules, textures, and reusable characters, presenters and series settings. The Studio supports English and Turkish. The audio pipeline covers narration in more than one language, captions, lip sync, music and sound effects. The output is a complete publishing package: subtitles, metadata, and a vertical cut for shorts.

The project needs Node.js 24, pnpm 11, FFmpeg, and either Claude Code signed in or an ANTHROPIC_API_KEY. The README names voice, image and video APIs as separate providers configured through .env. The episode budget caps the paid APIs. Claude agent usage is tracked on its own and can exceed the cap.

Who this is for

The audience is a content team, a solo founder or a one-person channel. Someone with a backlog of videos that never got made, because each one costs a week. The project does not claim machine-made video is better. It claims it is finished. A team can run a brief through it. They judge the output on their own work. They keep or throw away the result.

The README shows no video examples, no benchmark and no comparison with a human-made clip. A reader has to judge the output themselves. That is honest. It is also the limit. A team making a hero video for a product launch will still want a person at the camera. A team with 20 explainers that never get made is the right audience here.

Source

AI Video Factory on GitHub.

SourceGitHub

This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.

Share
LinkedInX
Start a project