Esta página também está em português. Ver em português

The method behind Kaneda. The first chapter is free.

Five chapters: plate first, the prep and remap sandwich, five prompt rules, the spending governor and how to measure. US$ 20 on its own, included in every Kaneda license.

Included in the full license and in the beta seats.

Want the code too? Get Kaneda and the method comes with it.

Free chapter

Plate first

Most failed takes fail at the start. The video model is asked to swap the person, the clothes, the props and the scene while it also copies the motion. Faces blur, hair changes and bystanders appear.

Plate-first splits the job. First you build one still of the finished scene, a plate, with an image model and your reference photos. You check the plate for free or for cents. Only then do you hand it to the video model as the image to animate.

In our approved takes the plate came first. In our control run, which went straight from a raw photo to video, the hairstyle was lost and a stranger's accessories showed up. C6 The side-by-side is on the proof page.

In the paid method

The sandwich: prep, model, remap

Why short clips make the model invent frames, and the three-step fix that returns the exact frame count and the original audio.

This chapter is part of the paid method.

In the paid method

Five prompt rules

The five sentences we add to every prompt, each with the failure that made us write it.

This chapter is part of the paid method.

In the paid method

A governor between you and the API

The five checks that run before a request leaves your machine, and what each one has saved us.

This chapter is part of the paid method.

In the paid method

Measuring, and where it failed

What to measure on every take, the metric that failed us, and how to label your own takes.

This chapter is part of the paid method.

Included in the full license and in the beta seats.

Want the code too? Get Kaneda and the method comes with it.