Questions before you buy.
Is it a subscription?
No. It is a one-time purchase. You pay once for the license. There is no subscription, no seat fee and no usage fee from us. You still pay your video provider (Higgsfield or another API) for the generations you run.
Can I buy only the method?
Yes. The written method costs US$ 20 on its own. It is included in the full license and in the beta seats, so you never pay for it twice.
What do I need to run it?
Python 3, ffmpeg and an API key from your video provider. Kaneda runs on your machine. The method, the prompt rules and the prep and remap step work with any video model; the governor ships with a client for Higgsfield's API. We never see your key or your media.
Is this affiliated with Higgsfield?
No. It is independent software that uses their API with your own credential. We use their product names only to say what Kaneda works with and to compare results.
Who has to consent to what I upload?
You do. If a real person's face or voice is in your inputs, you need their consent, and Higgsfield's terms ask you to confirm it. We do not supply faces, and this site shows no one from our team or from our customers.
What does Kaneda not do?
It does not generate video by itself or resell Higgsfield access.
It does not promise that a take will look right. It quotes, caps, measures and logs, and then a person approves.
Does Kaneda phone home?
Kaneda talks to two hosts: Higgsfield's API and, only if you add a Jev key, an optional decision API. There is no license server. Nothing is sent to us unless you switch the optional plane on. C25
How do model prices work?
What are the beta seats?
The first 100 buyers get Kaneda at US$ 97. In return we publish one case study with your recipe, and we ask for written consent from anyone who appears in it. When the seats are gone the offer closes.
Why publish where your own metric failed?
Because you will measure your own takes, and the number that looks best is not always the take you will approve. We would rather you know that before you buy.
Which models does it support?
Wan 3.0 reference-to-video is the main route and the one we have tested most. The executor also has routes for Genjutsu motion transfer and object swap and for Seedance, which we have used less. Everything except the API client is model-agnostic: with another provider you call its API yourself and still use prep, remap, the rules and the telemetry.