AI-facilitated, not AI-generated.

Where AI is used, and where it is not

You can edit every script

The script is the video: it decides what the narrator says and what the pictures show. You have full control of it: edit the narration, and say what you want to see at each point in the video. Nothing is rendered until you have approved it, and you can ask for changes as many times as your revision rounds allow. Your approval is the final check.

Human review adds a second person to that process. It is a check by an editor who was not involved in preparing the script. It is not a guarantee, and it does not replace your own reading of the script against what you know of the research.

A disclosed, synthetic voice

You choose the narrator from 28 high-quality synthetic voices in English: British and American, women and men, with different pitches and paces. Listen to each one on the order form. The voices run efficiently on our own machines.

The narration is not a recording of a person, and we never pretend it is. Every video carries a disclosure that the voice is synthetic and that the script was drafted with AI assistance: in the first caption, in the transcript and in the video file's own details. Our terms ask you to keep it when you publish.

Built, not generated

Generative AI video models paint every frame, so they can invent a number, a label or a result that sounds right but is not in the paper. We build each frame from the paper's own words and data instead, so the video can be checked against its source:

• No made-up facts: our software draws only what is in the script you approve, and the script notes the page, table or figure behind each line, so you can check it against the paper.
• Every word on screen is typeset, not painted by an image model, so there is no garbled lettering and no text that is not in the script you approved.
• Charts are plotted from the paper's own numbers. In our pilot video of Hubble's 1929 paper, all 24 nebulae come from his Table 1.
• Diagrams are drawn deliberately, in code, to show what the script says.
• What the video says is the script you approved, with human review by one of our editors if you add it.

That is a description of the method, not a promise that a video cannot contain a mistake: the script is where the words come from, which is why you approve it.

Why AI video models get text, numbers and charts wrong, what the model makers say about it, and how our process keeps a video true to the paper: how we keep videos accurate.

How many videos our engine has rendered, and the electricity we estimate that saved against AI video generation and a traditional animator: our numbers.

Respecting the paper's copyright

The research is the authors'; our job is to help it reach more people. A redrawn video copies no text or images from the paper: it relays the paper's ideas and findings in our own words and drawings, so it does not normally need the publisher's permission. A video that shows the paper's own pages reproduces the paper, so it needs the right to adapt it. You choose which when you order. Read more about copyright protection and use here.

No generated or scraped artwork

The pictures use no generated images and no scraped artwork, and imitate no one's visual style.

Your paper stays yours

Your paper is used to make your video and nothing else. We never publish a video, or use it as a sample, without your written permission, and papers you mark as confidential or embargoed get extra care. The privacy policy names the language model provider and what happens to your data.