# Auteuriq

> Auteuriq builds creative intelligence for people with a point of view. Its first model, Cadence v1, is an AI model trained only for video editing, paired with a native editing engine. It learns an editor's style from a few of their own edits and brings that taste to new footage. Cadence v1 is in private research preview; early access is by request at https://auteuriq.com/#early.

Key facts:

- Company: Auteuriq (https://auteuriq.com), contact hello@auteuriq.com
- Product: Cadence v1, an AI video editing model with its own editing engine
- Status: private research preview, opening to a small group of creators, editors and studios first
- Audience: creators, editors and studios who cut a lot of video and want to keep their own style
- Thesis: "The difference isn't the footage. It's the decisions." Cadence doesn't learn your presets. It learns your eye.

## What Cadence v1 does

- Watch: reviews every take for horizon, focus, exposure and motion, and finds the moment the subject is strongest.
- Select: sorts usable takes from the cutting-room floor and keeps every rejected take in case you disagree.
- Assemble: finds the story (hook, build, hold, release), removes dead air and opens on the strongest moment.
- Rhythm: cuts where the music breathes, lands downbeats on action and steps the score back for dialogue or sound.
- Time: speed ramps, freeze frames and reverse.
- Mood: colour grading, with live waveform and vectorscope to keep it honest.
- Draw: charts, diagrams and hand-drawn motion graphics, animated on twos and cut into the footage.
- Type: 3D titles that sit inside the shot, occluded by objects, lit by the scene and locked to the camera.
- Track: follows moving objects and keeps labels attached to them.
- Build: layered compositing, sky replacement and CG.
- Style: learns an editor's characteristic (average shot length, cuts on beat, transitions, grade) from a few of their edits.

## How it thinks

Cadence treats an edit as a program rather than pixels. Every cut, ramp, grade, title and track is a line the model writes, the editor opens and the engine renders. The program has five layers: Intent (what it should feel like, and for whom), Structure (what happens, in what order), Treatment (time, colour and sound as curves), Composition (type, graphics, masks and tracks) and Scene (depth, light, cameras and CG). A perception stack underneath watches motion, meaning, speech and music; the editor's learned style conditions every layer.

## The concept film

- [Characteristic](https://auteuriq.com/): a 13-chapter, scroll-driven film about a lighthouse that edits itself as you scroll. It is rendered live in the browser with WebGL and Web Audio, with no video files. Every decision in it was authored by the Auteuriq team to show where Cadence v1 is headed; it is not model output.

## Pages

- [Home and concept film](https://auteuriq.com/): the film, the editable program behind it, the five layers and the FAQ
- [Questions](https://auteuriq.com/#faq): what Cadence v1 is, what it does, who it's for and whether you can use it today
- [Early access](https://auteuriq.com/#early): request a seat in the research preview
- [Privacy notice](https://auteuriq.com/privacy): what the early-access list stores and how to have it deleted