Voice banking: preserving your own voice, and consent done right

The case for cloning, told straight

Most of this documentation set is about restraint: who must not be cloned, and why. This page is the reason the feature exists anyway. When the speaker and the consenting adult are the same person, a voice clone stops being a risk and becomes something close to a prosthetic: a way to keep sounding like yourself when your body stops cooperating, or to license your own instrument on your own terms.

Every rule in the other guides exists to protect exactly this use, not to compete with it.

Voice banking: recording yourself while you can

People diagnosed with ALS and similar conditions often lose speech as the disease progresses. Voice banking is the practice of recording yourself early so a synthetic version of your own voice, rather than a generic one, can speak for you through a communication device later. Message banking is its companion: saving real recordings of the phrases that matter, in your own delivery.

This is established clinical practice with real support behind it. Team Gleason, founded by NFL player Steve Gleason after his own diagnosis, helps people living with ALS in the US choose tools, set up recording, and fund the process. The ALS Association publishes guidance on voice preservation and walks through the options. Both say the same urgent thing: start early, while your voice is still fully yours, because clean recordings are the whole game.

Performers: licensing the replica instead of losing it

The other consent-done-right story is professional. After the 2023 strikes, SAG-AFTRA's AI agreements wrote digital replicas into union contracts: clear and conspicuous consent before a replica is made, disclosure of the intended use, and bargained compensation when it is used. A voice actor can now treat their voice as a licensable asset with terms, rather than something a buyer captures once and keeps forever.

The structure of those deals is worth copying even outside the union: consent that names the use, payment tied to the use, and the performer keeping the right to say no to the next one.

Why consent-first design is what enables this

A voice bank is only valuable if the voice stays under its owner's control: provably authorized at creation, private by default, and deletable for real. Those are the same properties that make abuse hard, which is the quiet point of this whole series: the engineering that protects a stranger from being cloned is the engineering that makes your own clone trustworthy enough to rely on. Platforms that skip the consent gate are not more convenient for the ALS patient or the working performer; they are less safe for both.

What this means on Cantari

You will be able to preserve your own voice here once cloning ships (it is coming soon). The create flow will take 10 to 20 seconds of clean speech (up to two minutes), offer prepared reading passages so you do not have to improvise, and require the consent attestation, which for your own voice is simply the truth. The clone will be private to your account, generations will land in your library, and deleting it will remove the voice, the engine-side model, and the stored reference clip.

Two honest notes for anyone banking a voice against a diagnosis. First, keep your original recordings somewhere you control as well; cloning here is coming soon, English will be the tested path, and your raw audio is the asset that outlives any platform, ours included. Second, record more than you think you need, in your natural registers: the advice every voice-banking program gives, because it is right.